Methods, systems and computer-readable media for assessing meat quality.

Description

1 / 45 “METHODS, SYSTEMS AND COMPUTER-READABLE MEDIA FOR EVALUATING MEAT QUALITY” TECHNICAL FIELD

[0001] The modalities described generally refer to methods, systems and computer-readable media for evaluating the meat quality of animal carcasses, such as chilled carcasses. BACKGROUND

[0002] In the field of meat quality assessment, carcasses are normally assessed by inspection to determine the quality of the meat they contain.

[0003] Existing techniques used to assess meat quality tend to involve capturing an image of an evaluation region of the carcass using a specific technical imaging device made to assess meat quality.

[0004] The aim is to solve or improve one or more deficiencies or disadvantages associated with this previous technique or, at least, to offer a useful alternative to it.

[0005] Throughout this descriptive report, the word “compreender”, or variations such as “compreende” or “compreendendo”, will be understood as implying the inclusion of a stated element, number or step, or group of elements, whole numbers or steps, but not the exclusion of any other element, number or step, or group of elements, whole numbers or steps.

[0006] Any discussion of documents, acts, materials, devices, articles or the like that has been included in this descriptive report should not be considered as an admission that any or all of these matters form part of the state of the art or were of common general knowledge in the field relevant to this disclosure, as they existed prior to the priority date of each of the appended claims. SUMMARY

[0007] Some modalities are related to a Petition 870250107908, dated 11 / 25 / 2025, page 9 / 58 2 / 45 method comprising: determining video data comprising a sequence of frames, wherein at least some of the frames represent a portion of a carcass to be evaluated for quality; for a first frame in the sequence of frames: a) determining a frame suitability score for evaluating a first feature; b) in response to the fact that the frame suitability score is greater than a threshold frame suitability score for the first feature, determining the frame as a suitable frame for evaluating the first feature; c) determining, by a segmentation model, a region of interest in the frame for evaluating the first feature; d) determining, by a first feature prediction model, a prediction score for the first feature based on the determined region of interest; e) determining a confidence rating of the evaluation for the first feature;In response to the determination that the confidence rating of the assessment for the first characteristic did not exceed the confidence limit of the assessment for the first characteristic, perform steps a) and e) for a subsequent frame in sequence; and in response to the determination that the confidence rating of the assessment for the first characteristic exceeded the confidence limit of the assessment for the first characteristic, determine a quality assessment measure for the first characteristic based on the prediction score for the first characteristic of the determined suitable frames.

[0008] The method may further comprise determining the confidence rating of the assessment for the first feature: determining one or more of: (i) a number of suitable frames determined for the first feature; and (ii) a function of the frame suitability scores for the first feature of the suitable frames determined.

[0009] The method may also include determining the confidence rating of the assessment for the first feature: determining a function of the prediction scores for the first feature of the determined suitable frames. In this case, determining a function of the prediction scores for the first feature of the suitable frames Petition 870250107908, dated 11 / 25 / 2025, page 10 / 58 3 / 45 determined may involve determining an arithmetic mean obtained in the interval: [Bmin, Bmax], where Bmin is defined by the function: Bmin= PB- (tolIQRx IQR)

[0010] and where Bmax is defined by the function: Bmax = PA + (tolIQR x IQR.)

[0011] where Pa is a prediction value of a first percentile of prediction scores for the first characteristic; where Pb is a prediction value of a second percentile of prediction scores for the first characteristic; where tolIQR is a tolerance parameter; and where IQR is an interquartile range defined by IQR = Pa - Pb.

[0012] The method may further comprise excluding the lowest pdrop percentile from the prediction scores for one or more of: (i) a measure of lack of brightness; (ii) a measure of sharpness; and (iii) a measure of segmentation size; where pdrop is a configurable parameter. The method may further comprise: determining a frame number limit for the frame sequence; in response to the determination that the number of frames exceeds the frame number limit, discarding the frame suitability score, the prediction score, and the confidence rating for the first determined frame. The first percentile may be a lower percentile than the second percentile.

[0013] The method may also comprise: in response to the determination of the frame as a suitable frame for evaluating the first characteristic, perform steps c) and d); and, in response to the determination of the frame as not being a suitable frame for evaluating the first characteristic, omit steps c) and d).

[0014] Determining video data may involve receiving a video stream.

[0015] The frame adequacy score may be based on one or more of: (i) a measure of lack of glare, (ii) a measure of sharpness, (iii) a measure of segmentation size; (iv) a measure of margin Petition 870250107908, dated 11 / 25 / 2025, page 11 / 58 4 / 45 of the segmentation of the region of interest; (v) a rounding measure of the segmentation of the region of interest.

[0016] Determining the adequacy score of the framework may involve determining the adequacy score of the framework according to the following equation: St= Στmáx(vra.t + b'., 0),

[0017] where Sté is the adequacy score; vré is a classification type r value for the first frame; a£ is a scaling coefficient of classification type r for feature t; eb$ is a displacement coefficient of classification r for feature t.

[0018] The method may also include: the output of the quality assessment measure of the first feature for a user interface.

[0019] The first characteristic may include any of the following: marbling; fineness of marbling; rib eye area; rib fat ratio; intramuscular fat; fat color; and meat color.

[0020] The method may further comprise: for the first frame in the sequence of frames: f) determining a frame adequacy score for evaluating a second feature, where the second feature is different from the first feature; g) responding to the fact that the frame adequacy score is greater than the limit frame adequacy score for the second feature, determining the frame as an adequate frame for evaluating the second feature; h) determining, by a second segmentation model, the region of interest in the frame for evaluating a second feature; i) determining, by a second feature prediction model, a prediction score for the second feature based on the determined region of interest; and j) determining an evaluation confidence index for the second feature; in response to the determination that the confidence index of Petition 870250107908, dated 11 / 25 / 2025, page 12 / 58 5 / 45 If the assessment for the second characteristic did not exceed the confidence limit for the second characteristic, perform steps a) and ae) for a subsequent frame in sequence; and in response to the determination that the confidence index for the second characteristic exceeded the confidence limit for the second characteristic, determine a quality assessment measure for the second characteristic based on the prediction scores for the second characteristic from the appropriate frames determined. The second characteristic may be different from the first characteristic and comprises any of the following: marbling, fineness of marbling, rib eye area, rib fat thickness, intramuscular fat, fat color, and meat color.

[0021] The frame adequacy score may be based on one or more frame adequacy ratings, each frame adequacy rating indicating the adequacy of a frame with respect to a specific measure, and the method also comprises, for each of the one or more frame adequacy ratings: the comparison of the frame adequacy rating with a respective specific rating threshold of the measure; and the response to the determination that the frame adequacy rating does not meet the threshold, the determination of the frame as inadequate and the exclusion of the frame from further processing.

[0022] In some embodiments, the step of determining, by a segmentation model, a region of interest in the frame to evaluate the first feature, includes: receiving a point cloud and isolating the area of ​​the point cloud corresponding to the region of interest; determining a plane, fitted to the points of the point cloud; projecting the points of the point cloud onto the plane to generate a plurality of projected planar points; rasterizing the projected planar points into a high-resolution image in which each pixel cell of the high-resolution image corresponds to a region of the plane of a real-world area; converting the high-resolution image into a binary image; applying a morphological operation to fill holes in the binary image; identifying the outline of the region of interest in the binary image; and, in response to the identification of only one Petition 870250107908, dated 11 / 25 / 2025, page 13 / 58 6 / 45 single contour, convert the area inside the contour into an estimate of the real-world area of ​​the region of interest.

[0023] Some embodiments relate to a meat evaluation system comprising: at least one processor; memory accessible to at least one processor and comprising computer executable instructions which, when executed by at least one processor, cause the system to perform the described method.

[0024] The system may further comprise: a user interface configured to display the quality assessment measure determined for the first characteristic. The system may also comprise a video capture device for capturing video data, controlled by at least one processor.

[0025] Some modes are related to a non-transient, machine-readable medium that stores instructions which, when executed by one or more processors, cause an electronic device to perform the described method. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a schematic example of an animal product subjected to meat quality assessment, according to some methods;

[0027] Figure 2 is a block diagram of an analysis device for performing meat quality assessment, according to some methods;

[0028] Figure 3 is a flowchart of a method for evaluating meat quality, according to some modalities; and

[0029] Figures 4A and 4b are examples of screenshots of user interface displays of the analysis device in Figure 2, according to some embodiments. DETAILED DESCRIPTION

[0030] The modalities described generally refer to Petition 870250107908, dated 11 / 25 / 2025, p. 14 / 58 7 / 45 methods, systems and computer-readable media for evaluating meat quality from animal carcasses, such as chilled carcasses.

[0031] In meat processing environments, it is often advantageous to quickly and reliably classify meat carcasses for further processing. Figure 1 is a schematic example of a meat product 140 subjected to meat quality assessment using an analysis device 110, according to some embodiments. The analysis device 110 is configured to assess the meat product 140 to determine a quality assessment of one or more traits or characteristics of the meat product 140. For example, characteristics may include one or more of: marbling, fineness of marbling, rib eye area, rib fat thickness, intramuscular fat ratio, meat color, fat color, and eye muscle area (EMA). The eye muscle area may be determined as a measure of the rib eye muscle in square units, for example, cm2.Rib fat thickness may refer to the actual absolute thickness of the rib fat in the meat product 140. In some embodiments, rib fat thickness may include a proportional value compared to another aspect of the geometry of a cut surface of the meat product 140. In some embodiments, marbling fineness may include a measure of how small and / or uniformly distributed the fat particles are within the EMA. Intramuscular fat ratio may also be referred to as intramuscular fat or IMF.

[0032] In general, in the embodiments described, a user using the analysis device 110 captures video data of the meat product 140 using the analysis device 110. For example, the analysis device 110 may comprise a video capture component 112. The video data comprises a sequence of frames 104, each frame representing a view or image of the meat product 140. The analysis conducted by the analysis device 110 may be performed simultaneously, or nearly in real time, with the capture of the video data.

[0033] Analysis device 110 is configured for Petition 870250107908, dated 11 / 25 / 2025, page 15 / 58 8 / 45 Analyze frames of the video data to determine quality assessment measures for the feature(s).

[0034] The quality assessment measure(s) for the characteristic(s) can be determined using machine learning (ML) models, such as neural networks. The model(s) determine(s) the prediction score(s) for the respective characteristic(s) for each frame from a plurality of captured frames. A frame suitability score can be determined for each frame to determine if the frame is suitable for use in meat quality assessment. In some embodiments, the determination of a frame's suitability and the determination of the prediction score(s) for the respective frame characteristic(s) are performed in parallel or virtually simultaneously.

[0035] Analysis device 110 can determine a confidence rating of the assessment to determine whether a quality assessment measure should be determined based on the acquired information. For example, the confidence rating of the assessment may depend on whether a sufficient number of frames have been acquired (e.g., at least ten) and / or whether a sufficient number of suitable frames have been acquired and / or whether a function of the prediction scores for the suitable frames meets a threshold, e.g., being sufficiently high or sufficiently low.

[0036] After the analysis device 110 determines the quality assessment measure(s) for the characteristic(s), the analysis device 110 provides or sends the determined quality assessment measure(s) to a user interface 120 of the analysis device 110. For example, the analysis device 110 may display the determined quality assessment measure(s) on a display screen 130 of the user interface 120, thus allowing a user to see the determined quality measure of the meat product 140 in relation to each characteristic. The analysis device 110 may store the determined quality assessment measure(s) in memory 210. The analysis device 110 may send determined quality assessment measures to a Petition 870250107908, dated 11 / 25 / 2025, page 16 / 58 9 / 45 external device via communication interface 226.

[0037] Examples of screenshots of a display 130 of the user interface 120 of the analysis device 110 are shown in Figures 4A to 4b, as discussed in more detail below.

[0038] Performing meat quality assessment of meat products 140 using video can allow for a high degree of accuracy and / or efficiency. As the assessment method is performed on a series of frames 104, data can be acquired relatively quickly and / or the determination of desired characteristics can be more precise, as it is determined from a matrix or plurality of frames 104. This may be preferable to the use of single or individual images, for example, since single images may be limited by factors such as ambient lighting, lack of sharpness, brightness and / or other factors that may affect the ability to make an accurate assessment of meat quality from a single image. Furthermore, to capture a single high-quality image, it is often necessary to pause the operation of a meat processing line, which can contribute to costly delays in terms of refrigeration time and / or processing production.On the other hand, performing an evaluation on more than one frame of an image or video set can offer the advantage of allowing operators to assess meat quality under a variety of lighting and / or image conditions without significantly affecting the accuracy and / or efficiency of the evaluation.

[0039] The described method can provide an efficient and / or reliable technique for evaluating meat quality. The ability to evaluate a video stream, comprising a sequence of images or frames 104, means that a relatively large amount of data can be captured in a short period of time for analysis. Determining frame adequacy on an individual or “frame-by-frame” basis means that operators may not need special training to operate a camera, for example, to accurately position a camera for image capture, as may be the case when evaluation is performed on single images. In this way, the benefits of using Petition 870250107908, dated 11 / 25 / 2025, page 17 / 58 10 / 45 trained machine learning models from the Meat Quality Assessment Module 214 mean that operators can capture videos of the general area and rely on the functionality of the Meat Quality Assessment Module 214 to evaluate which parts of the video are unsuitable for analysis. This approach can be very advantageous in a slaughterhouse, for example, where environmental conditions such as ambient lighting cannot be controlled and, in some cases, carcasses may be moving while being graded. Furthermore, the systems and methods described can be particularly beneficial for other processing plants, which can benefit from the reliable and / or efficient determination of meat quality in products they receive for further processing. The systems and methods described can be particularly beneficial for retailers, to identify meat quality at an individual cut level.Furthermore, the systems and methods described can be useful in consumer applications to allow for easy-to-use classification of purchased meat products, assessment of meat product value, and / or to help evaluate ideal cooking times and / or procedures.

[0040] Figure 2 shows the analysis device 110, according to some embodiments. As illustrated, the analysis device 110 may comprise a video capture device 112 for capturing video data of a meat product to be evaluated. The analysis device 110 may comprise a user interface 120 for outputting, to a user, for example, certain quality assessment measure(s) for the characteristic(s) of the meat product. The analysis device 110 may be fixed to a support 115.

[0041] In some forms, the analysis device 110 could be a smartphone, such as a Samsung Galaxy™ model smartphone. In these cases, the video capture device 112 could comprise a digital camera integrated into the smartphone. In other embodiments, the video capture device 112 could be a separate unit installed remotely from the controller 202. In these cases, the video capture device 112 could comprise an independent video camera. Petition 870250107908, dated 11 / 25 / 2025, page 18 / 58 11 / 45

[0042] In some embodiments, the video capture device 112 may comprise a 2D camera configured to capture video. The video capture device 112 may be coupled to an enclosure, such as a metal enclosure, configured to fix or hold the video capture device 112 in a specific position so that it captures video from a fixed angle and / or orientation. The fixed angle and / or orientation may be selected to ensure that the video capture device 112 captures an appropriate or adequate view of the meat product. However, it is important to note that the use of a cover is optional and not required.

[0043] In some embodiments, the video capture device 112 may comprise a 3D camera configured to capture video data with depth information. In these embodiments, the 3D camera may comprise an Intel REALSENSET™ camera or an equivalent camera.

[0044] This 3D camera can be configured to capture frames of the meat carcass from various different angles and / or orientations. The use of a 3D camera can therefore eliminate the need for a cover to maintain a fixed angle and / or orientation, as mentioned above, and, in some embodiments, for a static device 115. For example, the video capture device 112 can be deployed on a robotic arm (not shown), which can assist in acquiring frames at a variety of angles and / or orientations. This can be particularly useful in meat processing conditions, which may involve moving carcasses on a chain, as enabling a wider range of angles and / or orientations can mean that proper frame acquisition and / or accurate meat assessment can be reliably maintained as carcasses move along the chain.

[0045] Furthermore, the use of a 3D camera for the video capture device 112 may not only allow the capture of a series of frames from different angles and / or orientations, but may also allow the capture of relatively high-resolution images and, consequently, a greater amount of information than a 2D camera might allow. This may Petition 870250107908, dated 11 / 25 / 2025, page 19 / 58 12 / 45 improve the accuracy of subsequent processes, such as selecting the genetic stock and determining appropriate treatments for the stock based on the evaluated characteristics.

[0046] In some modalities, when at least one of the evaluated or measured characteristics is eye muscle area (EMA), marbling, rib eye area, rib fat thickness, meat color, fat color, marbling fineness, and / or intramuscular fat ratio, the use of a 3D camera can eliminate the need for a grid, such as a point grid (e.g., a point planimeter), to be placed or overlaid on a region of interest of the carcass, such as the rib eye area, to allow for area determination. The use of such a grid can be cumbersome and inefficient, requiring operator skill and time for placement, as well as sanitary practices to be implemented to ensure the grid is kept clean and that there is intervention when moving it from one carcass to another. The use of a 3D camera can be particularly advantageous when EMA is only one of several different characteristics being evaluated.In this case, it may be necessary to capture at least two separate video streams: one with the grid of points in the region of interest of the carcass and another without the grid of points (which is not necessary or is not suitable for evaluating other features).

[0047] In some embodiments, the video capture device 112 captures a video stream to better enable near real-time processing and reduce the impact of storage on memory.

[0048] As illustrated in Figure 2, the analysis device 110 comprises a controller 202 configured to perform the analysis and output the results. For this purpose, the controller 202 may be in communication with the video capture device 112 and / or the user interface 120. The controller 202 comprises one or more processors 205 in communication with the memory 210.

[0049] Processor(s) 205 may be arranged to retrieve data from memory 210 and execute program code stored in memory 210 to perform quality assessment functionality. Petition 870250107908, dated 11 / 25 / 2025, page 20 / 58 13 / 45 described. The processor(s) 205 may include more than one electronic processing device and / or additional processing circuits. For example, the processor(s) 205 may include multiple processing chips, a digital signal processor (DSP), analog-to-digital or digital-to-analog conversion circuits, and / or other processing circuits or chips that have processing capabilities to perform the functions described in this document. The processor(s) 205 may perform all the processing functions described in this document locally on the analysis device 110.

[0050] Memory 210 may comprise a UI module 212 which, when executed by processor 205, sends and receives instructions from user interface 120 and allows the output, as visual display and / or audio output, of information stored in memory 210 on user interface 120.

[0051] Memory 210 comprises a meat quality assessment module 214 which, when executed by the processor(s) 205, determines the meat quality assessment measure(s) for the respective characteristic(s) of a meat product. In some embodiments, the meat quality assessment module 214 receives video data comprising a sequence of frames 104 as input. For example, at least some of the frames 104 may represent a portion of a carcass to be assessed for quality. The quality assessment module 214 may generate a determined quality assessment measure for one or more characteristics related to meat quality.

[0052] As mentioned above, the video capture device 112 may comprise a 2D camera or a 3D camera. In embodiments where the video capture device 112 comprises a 3D camera, the video capture device 112 may be external to or distinct from the analysis device 110. The memory 210 may comprise a 3D camera interface module 213. The 3D camera interface module 213 may comprise code libraries that, when executed by the processor(s) 205, cause the 3D camera interface module 213 to send instructions to and receive instructions from the 3D camera. Petition 870250107908, dated 11 / 25 / 2025, page 21 / 58 14 / 45 instructions from her. In modalities that utilize the depth functions of a 3D camera, such as in the evaluation of characteristics like EMA, marbling, rib eye area, rib fat thickness, meat color, fat color, marbling fineness and / or intramuscular fat ratio, the processing of the 3D camera depth data can be performed by the camera interface module 213 and / or the meat quality evaluation module 214. In some modalities, the 2D camera can be used to evaluate characteristics such as EMA, marbling, rib eye area, rib fat thickness, meat color, fat color, marbling fineness and / or intramuscular fat ratio. In some modalities, the video data obtained by the 2D camera can be evaluated using analysis modes without depth.In some applications, video data obtained by the 2D camera can be evaluated using monocular depth estimation, stereo depth estimation from two different lenses, or similar techniques to perform a depth-based analysis using the 2D image(s) as the data source.

[0053] The meat quality assessment module 214 comprises an object determination module 216 and an object quality determination module 218.

[0054] The object determination module 216 may comprise program code that, when executed by the processor 205, determines an area or region of interest 132 in a frame of the video data. Multiple areas or regions of interest 132 may be determined by the object determination module 216 for any frame 104. For example, a different or distinct area or region of interest 132 may be determined for the evaluation of each feature.

[0055] In some embodiments, the object determination module 216 may comprise one or more segmentation models 220. Each segmentation model 220 may be configured to determine an area or region of interest 132 within a frame for analysis. In some embodiments, a given area or region of interest may be Petition 870250107908, dated 11 / 25 / 2025, page 22 / 58 15 / 45 suitable for performing a meat quality assessment measure for more than one, or even all, characteristics of interest. The segmentation model(s) 220 may be a machine learning model, trained on a dataset of processed carcass images and / or meat products, and configured to determine the presence, boundaries, and / or segmentation of the meat product in a frame 104.

[0056] The 220 segmentation model may comprise a convolutional neural network. The 220 segmentation model may be configured to identify the region of interest 132 in a frame 104. After the region of interest 132 is identified, the 220 segmentation model may crop frame 104 to focus on the proportions of the area of ​​interest 132. In some cases, other parts of the image are masked and the cropped (and / or masked) frame 104 may be scaled. The cropped and / or masked and / or scaled frame 104 may be fed into an additional convolutional network within the 220 segmentation model to produce a scalar feature prediction.

[0057] When the characteristic being evaluated is EMA, the 220 segmentation model can be configured to process frames with a planimeter positioned or placed in the area of ​​interest on the carcass, such as the rib eye area. In some embodiments, the 220 segmentation model can be configured to operate according to a first state or mode of operation to evaluate characteristics other than EMA that do not require placing a grid of points in the area of ​​interest on the carcass when acquiring video frames, such as marbling, rib eye area, rib fat thickness, meat color, fat color, marbling fineness, or intramuscular fat ratio.The 220 segmentation model can be configured to operate according to a second state or mode of operation to evaluate characteristics such as EMA, marbling, rib eye area, rib fat thickness, meat color, fat color, marbling fineness, or intramuscular fat ratio, and which require placing a grid of points on the area of ​​interest of the carcass when acquiring the video frames, such as, for example, when a 2D camera is used. Petition 870250107908, dated 11 / 25 / 2025, page 23 / 58 16 / 45 being used. User interface 120 can be configured to allow the user to choose a desired operating mode for segmentation model 220. For example, user interface 120 can be configured to present or display a user interface element to allow a user to select a desired operating mode for segmentation model 220.

[0058] For the second mode of operation, the segmentation model 220 can be trained on images of meat products with grids to identify the eye muscle and recognize the presence of planimetric points in frame 104.

[0059] In some embodiments, the object determination module 216 may comprise a MobileNetV3 encoder with a Lite Reduced Atrous Spatial Pyramid Pooling segmentation decoder. In other embodiments, the object determination module 216 may comprise other machine learning models such as DeepLabv3, Mask R-CNN, EARNet, etc. In some embodiments, a binary cross-entropy loss function is used. However, it will be considered that other loss functions, such as data loss or sparse categorical cross-entropy, may be used. In some embodiments, training data augmentation techniques may be used to augment or adjust the image, such as: rotation and / or inversion and / or contrast adjustment and / or brightness adjustment and / or cropping and / or resizing of the image.

[0060] In some modalities, the 220 segmentation model, when performing an EMA segmentation process, may comprise YOLOv4-tiny. The loss function used may be a generalized IoU loss model.

[0061] The object quality determination module 218 may comprise program code that defines one or more feature prediction models which, when executed by processor 205, determine a prediction score for each respective feature based on the area or region of interest, as determined by the object determination module 216. The object quality determination module 216 may Petition 870250107908, dated 11 / 25 / 2025, page 24 / 58 17 / 45 understand a machine learning model, trained on a dataset of images of processed carcasses and / or meat products, and configured to determine characteristics such as marbling, marbling fineness, rib eye area, rib fat thickness, intramuscular fat, fat color, and meat color in images representing meat products.

[0062] In some embodiments, the object quality determination module 216 may comprise a MobileNetV3 encoder with custom pooling. The encoder may utilize fully connected layers. In other embodiments, the object quality determination module 216 may comprise an EfficientNetv2 model. In other embodiments, the object quality determination module 216 may comprise an FBNet model. In other embodiments, the object quality determination module 216 may comprise a ShuffleNetv2 model. The loss function used in the object quality determination module 216 may comprise a mean absolute error function. In other embodiments, the loss function may be a root mean square error function.

[0063] In some embodiments, the training data for the machine learning models used in the object quality determination module 216 can be augmented by rotating the input data. In some embodiments, the training data for the object quality determination module 216 can be augmented by inverting the input data. In some embodiments, the training data for the object quality determination module 216 can be augmented by changing the contrast of the input data. In some embodiments, the training data for the object quality determination module 216 can be augmented by changing the brightness of the input data. In some embodiments, the training data for the object quality determination module 216 can be augmented by changing the brightness of the input data.In some modes, the training data for the object quality determination module 216 can be augmented by clipping the input data. In some modes, the data of... Petition 870250107908, dated 11 / 25 / 2025, page 25 / 58 18 / 45 training of the object quality determination module 216 can be increased by resizing the input data. Scaling the data in this way can create a more resilient and accurate dataset.

[0064] The meat quality assessment module 214 comprises a frame adequacy determination module 224. The frame adequacy determination module 224 can be configured to determine whether a frame acquired from video data is adequate or sufficient to assess a specific characteristic of the meat product 140. In some embodiments, the frame adequacy determination module 224 is configured to determine a frame adequacy score to predict each respective characteristic. Some frames may be adequate or sufficient to assess a first characteristic in a meat product 140, but may be insufficient or inadequate to assess a second characteristic in the meat product 140. Thus, the frame adequacy determination module 224 may apply different criteria and / or thresholds in determining frame adequacy scores for the assessment of different characteristics.Similarly, the 224 frame adequacy determination module can apply the same criteria and / or limits in determining frame adequacy scores for evaluating different characteristics. In some embodiments, the 224 frame adequacy determination module can compare the frame adequacy score for a characteristic with a threshold frame adequacy score for the characteristic and, in response to a frame adequacy score being higher than the threshold frame adequacy score for the characteristic, the 224 frame adequacy determination module can determine the frame as an adequate frame for evaluating or predicting the characteristic. In response to a frame adequacy score being lower than the threshold frame adequacy score for the characteristic, the 224 frame adequacy determination module can determine the frame as an inadequate frame for predicting the characteristic.In some modalities, the module for determining the adequacy of table 224 may discard or discount the inadequate table for evaluation. Petition 870250107908, dated 11 / 25 / 2025, page 26 / 58 19 / 45 posterior. However, it should be considered that a frame deemed unacceptable or inadequate for evaluating a first characteristic may, nevertheless, be adequate for evaluating a second characteristic. In some modalities, a frame considered unacceptable for a first characteristic, but adequate for a second characteristic, may be discarded based on the fact that it is unacceptable for the first characteristic. The threshold adequacy scores may be configurable. The threshold adequacy scores may be configured differently for each characteristic. For example, each characteristic may have specific threshold values ​​for each characteristic; a frame that may be adequate for evaluating a first characteristic may be inadequate for evaluating a second characteristic.

[0065] The Meat Quality Assessment Module 214 can be configured to determine an assessment confidence rating for each trait. The assessment confidence rating can be based on the adequacy and / or number of adequate frames acquired and / or prediction scores determined for the adequate frames. In some embodiments, the Meat Quality Assessment Module 214 determines the assessment confidence rating for a trait by determining one or more of: (i) a number of adequate frames determined for the trait; and (ii) a function of the frame adequacy scores for the trait from the determined adequate frames. In some embodiments, the Meat Quality Assessment Module 214 determines the assessment confidence rating for a trait by determining a function of the prediction scores for the first trait from the determined adequate frames.

[0066] The meat quality assessment module 214 can determine whether to determine or generate a quality assessment measure for the characteristic(s) based on the respective assessment confidence rating(s). For example, the meat quality assessment module 214 can compare the assessment confidence rating with a threshold value for the respective characteristic. If the assessment confidence index exceeds the threshold Petition 870250107908, dated 11 / 25 / 2025, page 27 / 58 20 / 45 relevant, the meat quality assessment module 214 may determine or generate the quality assessment measure for the trait. The trait quality assessment measure may be based on the output of the object quality determination module 218, such as the prediction score(s) for the trait(s) of the appropriate frames determined. However, if the confidence index of the assessment does not exceed the relevant limit, the meat quality assessment module 214 may acquire and evaluate other frames of the video data for analysis.

[0067] The meat quality assessment module 214 can be configured to determine a carcass identifier. For example, the carcass identifier can be provided in the image frames 104. The carcass identifier can be on the meat product 140. The carcass identifier can be in an area near the meat product 140. The carcass identifier can be one or more of the following: a one-dimensional barcode; a two-dimensional barcode (such as a QR code or a DotMatrix system, for example); an alphanumeric code; a near-field communication (NFC) tag; an RFID tag. In some embodiments, the carcass identifier can be entered via the user interface 120. In these cases, the user interface receives the carcass identifier and transmits it to memory 210 via the processor 205.In some embodiments, the meat quality assessment module 214 sends a request to the processor 205 to retrieve a carcass identifier from an external application, using an application programming interface (API). The carcass identifier can be stored in memory 210. The carcass identifier can be stored in the meat quality assessment module 214. The meat quality assessment module 214 can associate the carcass identifier with an image frame 104. The meat quality assessment module 214 can associate the carcass identifier with a meat product 140. The use of a carcass identifier can allow for the precise retrieval of meat quality assessments that correspond to specific carcasses. The processor 205 can determine the carcass identifier through a programming interface. Petition 870250107908, dated 11 / 25 / 2025, page 28 / 58 21 / 45 of applications (API). The API may allow communication with a network of processing facilities where a carcass identifier is stored. The user interface 120 may comprise a display 130, configured to display the determined meat quality assessment measure(s). The user interface 120 may comprise a keyboard, a touch screen and / or a button-based interface. The display 130 may consist of an LED or LCD screen, such as the touch screen of a smartphone.

[0068] In some forms, the analysis device 110 can be mounted on a robotic arm (not shown). In these embodiments, the robotic arm (not shown) can be controlled by the controller 202 to orient the video capture device 112 of the analysis device 110 to match, align, or accommodate predetermined viewing angles of moving carcasses in a meat processing environment, for example, to allow consistent and / or relatively rapid capture of video images by the analysis device 110. The robotic arm (not shown) can improve the accuracy and / or efficiency of the analysis process, as a robotic arm mount may be able to better match the speed and orientation of the carcasses as they move during processing and / or maintain a consistent perspective. Furthermore, these embodiments can allow for a reduction in labor costs for factories.

[0069] In some forms, the analysis device 110 can be installed as part of a meat slicing assembly (not shown). In these embodiments, the video capture device 112 of the analysis device 110 can be oriented to capture video images of one face of a primary cut of meat. The output of the analysis device 110 can be used to determine the coating quality of the primary cut of meat. After determining the quality of the end of the primary cut of meat, it can be sliced ​​by the slicing assembly (not shown) and directed along one or more different conveyors. For example, each conveyor can lead to a separate or different processing area where the slices are processed. Petition 870250107908, dated 11 / 25 / 2025, page 29 / 58 22 / 45 labeled and / or packaged according to the assessed meat quality. In this way, these methods allow for relatively fast processing of meat products and / or a relatively greater specificity of meat quality throughout the primary cut of meat. This represents an improvement over existing high-volume meat processing solutions, which typically assign a single meat quality grade to processed primary cuts of meat – which may not capture higher quality portions that exist throughout the primary cut.

[0070] Communication interface 226 is accessible by processor 205 and configured to allow the exchange of information with devices external to the analysis device 110. Communication interface 226 may include components to receive a SIM card to facilitate communication via 2G, GSM, EDGE, CDMA, EVDO, 3G, GPRS, 4G, 5G or other suitable telecommunications networks. Communication interface 226 may include an Ethernet port to allow wired communication. Communication interface 226 may include a wireless Internet interface. Communication interface 226 may include a Bluetooth interface. Communication interface 226 may include one or more of the modes described above. Communication interface 226 may allow communication via an API. Communication interface 226 may allow communication via the Hypertext Transfer Protocol (HTTP, HTTPS).

[0071] Figure 3 is a flowchart of a method 300 for evaluating meat quality, according to some modalities. The method 300 can be performed by the analysis device 110 which executes the program code in memory 210, as the meat quality evaluation module 214.

[0072] In 302, the analysis device 110 determines the video data. The video data comprises a sequence of frames 104 or a series of images. At least some of the frames represent a part of a carcass or meat product 140 to be evaluated for quality. In some embodiments, the video data are captured by the video capture device 112 and provided to the meat quality assessment module 214. The sequence of Petition 870250107908, dated 11 / 25 / 2025, page 30 / 58 23 / 45 frames 104 can be stored in memory 210 for evaluation by the meat quality assessment module 214. The sequence of frames 104 can be stored temporarily and in near real-time during method 300, to avoid retaining large image files that could negatively affect storage in memory 210. In this way, in some modalities, each frame 104 of the video data can be analyzed as it is received and while the video data is still being captured by the video capture device 112.

[0073] In some embodiments, the characteristic being evaluated is the EMA. However, it is important to emphasize that the characteristic being evaluated may include one or more of the following: EMA, marbling, rib eye area, rib fat thickness, meat color, fat color, marbling fineness, or intramuscular fat ratio. As mentioned above, the analysis device 110 can be configured to operate in a second operating mode, for example, as selected by a user who provides input to the user interface or automatically upon detection of the placement of a grid in the carcass region being captured. The second operating mode can be specifically configured to evaluate the EMA of a meat product 140 with a grid placed over it, so that the frames of the meat product 140 captured by the analysis device 110 also represent the grid.The grid can be a planimeter, such as a translucent or substantially translucent point planimeter, for example. The grid allows the analysis device 110 to determine an area of ​​the carcass region of interest.

[0074] From 304 to 312, the meat quality assessment module 214 performs a series of actions for a first image frame 104 of the frame sequence.

[0075] In 304, the meat quality assessment module 214 determines a frame adequacy score to assess or predict a first characteristic. The frame adequacy score may relate to the suitability or sufficiency of a specific or candidate frame for Petition 870250107908, dated 11 / 25 / 2025, page 31 / 58 24 / 45 evaluates a specific characteristic. For example, scores with higher values ​​may indicate a greater degree of suitability for accurately predicting that characteristic. The adequacy score may be based on one or more classifications of the framework's adequacy.

[0076] In some modalities, the frame adequacy score may depend on a frame adequacy rating indicative of the absence of glare. For example, the Meat Quality Assessment Module 214 can be configured to convert the frame to grayscale. After the frame is converted to grayscale, the Meat Quality Assessment Module 214 can determine or calculate a number of pixels that exceeds a configurable glare value, as a proportion of the total number of pixels in the image. If the number of grayscale pixels exceeds the configurable glare value (frame adequacy rating threshold for absence of glare), the frame may be considered to have too much glare to be used in meat quality assessment. Such frames may be discarded or excluded from further analysis.

[0077] In some modalities, the frame adequacy score may depend on a frame adequacy rating indicative of image sharpness. For example, meat quality assessment module 214 may be configured to run a Fast Fourier Transform (FFT)-based method, configured to measure the high-frequency components of the image. If the high-frequency components of the image are determined to correspond to an image with sufficient sharpness (frame adequacy rating threshold for sharpness), the frame may be used in meat quality assessment. Frames that lack sufficient sharpness may be considered too blurry to be accurately assessed for meat quality and may be discarded or disregarded for further analysis.

[0078] In some modalities, the frame adequacy score may depend on a frame adequacy rating indicative of the segmentation size. This rating measures the size of the region. Petition 870250107908, dated 11 / 25 / 2025, page 32 / 58 25 / 45 of interest 132 identified in relation to the size of the entire frame. The meat quality assessment module 214, by the segmentation model 220, can determine the size of the identified segment as a number of pixels, as a segment pixel score. If the segment pixel score is below a threshold, the frame may be considered to have insufficient segmentation size and may be discarded or disregarded for further analysis. The frame adequacy classification threshold for segmentation size may be a minimum ratio of the determined segment pixel score compared to the total number of pixels in the frame.

[0079] In some modalities, the frame adequacy score may depend on a frame adequacy rating indicative of the segmentation margin. The segmentation margin is a rating related to the shortest distance between a segmented region of interest 132 and any edge of the frame. Very small values ​​may indicate that the region of interest 132 is being cut off by the edge of the frame. If a frame is determined by the meat quality assessment module 214 to have a segmentation margin score below a threshold (frame adequacy rating threshold for segmentation margin), the frame may be discarded or disregarded for further analysis.

[0080] In some modalities, the frame adequacy score may depend on a frame adequacy rating indicative of segmentation roundness. The segmentation margin is a rating that measures the roundness of the region of interest 132. This can be performed by the meat quality assessment module 214, fitting an ellipse to the area and obtaining its aspect ratio. Relatively highly oblique angles relative to the animal's cut surface may be considered undesirable for predicting some characteristics, and this value can be used as an indicator of the non-obliqueness of the camera angle.

[0081] In some modalities, the classification of the adequacy of the framework may include the application or output of an adequacy model. Petition 870250107908, dated 11 / 25 / 2025, page 33 / 58 26 / 45 of the frame or a frame quality model. The frame adequacy model may include a neural network configured to receive the input of a frame image. The neural network may be a convolutional neural network. The frame image may be segmented or unsegmented. In some modalities, the frame image may be resized or not resized. The frame adequacy model is configured to evaluate the quality or adequacy of the frame and generate a representative indicator. The representative indicator may be an indicator representing overall quality or adequacy, a quality or adequacy rating, or different aspects of quality or adequacy. In some modalities, the frame adequacy model is configured to generate a plurality of representative indicators. The representative indicator may be numerical or categorical.In some modalities, the assessment module 214 can be configured to discard frames based on the result of the frame adequacy model.

[0082] In some embodiments, the frame adequacy model may utilize other frame adequacy classifications, including those discussed in this document, as inputs to provide an assessment of quality or adequacy, such as frame adequacy classifications indicating segmentation roundness, segmentation margin, segmentation size, image sharpness, lack of glare, and the like. In some embodiments, the frame adequacy model may be used in combination with other frame adequacy measures to identify or determine adequate or high-quality frames. For example, frame adequacy measures may relate to the quality or content of the frame image, depending on the use of the frame adequacy model. In some embodiments, the frame adequacy model may be used in combination with or alongside frame feature prediction results.For example, to select a specific frame (or frames) from the plurality of frames in the video data for further processing or visual display.

[0083] For each animal characteristic, the relevance of Petition 870250107908, dated 11 / 25 / 2025, page 34 / 58 27 / 45 Each of the classifications described above varies, with, for example, brightness being extremely detrimental to some features but less impactful for others. Thus, for each feature, specific feature thresholds can be set for each of the different classifications, where frames are considered unsuitable for accurate prediction of that feature and excluded from further processing if any of these thresholds are not met.

[0084] In some modalities, one or more classifications may be used to calculate an overall score or the frame adequacy score for each frame with respect to each characteristic.

[0085] In some modalities, the St-frame adequacy score of a frame for a specific feature t can be calculated as: St Σ max(vra'r+ h,-, 0) r

[0086] where: Vr is a classification value of type r, a£ is a scaling coefficient of type r for the feature t, tf is a displacement coefficient of type r for the feature t (where tf and tf are constants between frames). For each feature / classification type pair (t, r), the coefficients tf and tf can be chosen so that higher values ​​indicate greater suitability for accurate forecasting, and St is always non-negative. The scaling and displacement coefficients for each classification type can be specifically selected or configured for each feature. This allows greater weight to be given to one or more classification types than to another(s), thus allowing the application of different criteria when considering the suitability of a frame to evaluate two different types of features.

[0087] In some modalities, a multiplicative approach may be adopted. For example, the St-frame adequacy score of a frame for a specific feature t can be calculated as: Petition 870250107908, dated 11 / 25 / 2025, p. 35 / 58 28 / 45 St^Π max(vrar + bj., 0) r

[0088] In some modalities, the St-frame adequacy score of a frame for a specific feature t can be calculated as the geometric mean of the components max(vra^. + b'r, 0).

[0089] In some modalities, the St-frame adequacy score of a frame for a specific feature t is calculated as the minimum or maximum of the individual components of max(vra\- + bj., 0).

[0090] In some modalities, the St-frame adequacy score of a frame for a specific feature t is calculated as the median of the components max(vra^. + b / ., 0).

[0091] In some modalities, the individual classifications vrart+ brtcan be included in any of the above modalities to calculate the St frame adequacy score for a 104 frame, for a specific feature t, without first determining the maximum of each classification with zero to produce a non-negative value.

[0092] In 306, in response to the fact that the frame adequacy score is greater than a limit frame adequacy score, the meat quality assessment module 214 determines that the frame is adequate for assessing or predicting the first characteristic. The limit can be set according to the characteristic of interest, as different characteristics may have different requirements for frame 104 to be determined as adequate for assessing or predicting the characteristic. When operating in the second mode of operation to assess EMA, the analysis device 110 can determine a frame 104 as being adequate for EMA assessment by determining whether a grid of points is represented in frame 104. In some embodiments, the analysis device 110 can determine that a frame 104 is adequate for EMA assessment if the meat quality assessment module 214 detects an eye muscle in frame 104. In some embodiments, the Petition 870250107908, dated 11 / 25 / 2025, page 36 / 58 29 / 45 analysis device 110 can determine that a frame 104 is suitable for EMA assessment by determining the presence of a grid of dots and an eye muscle in frame 104.

[0093] In 308, the meat quality assessment module 214 (e.g., the object determination module 216 or the segmentation model 220) determines an area or region of interest 132 in the frame 104, to assess or predict the first characteristic. For example, a frame of video data can be provided to the object determination module 216 as an input, and the object determination module 216 can provide, as output, one or more numerical values ​​indicative of the area of ​​the region of interest. In some embodiments, the object determination module 216 can generate an altered image or frame representing the region or area of ​​interest 132 in the frame, for example, as a delineated or bounded region. An example of the altered frame is illustrated in Figures 4a and 4b.

[0094] When operating in the second operating mode to evaluate the EMA, the analysis device 110 can determine the region of interest 132 within frame 104 by determining an eye muscle boundary within frame 104. In some modalities, the determination of the region of interest 132 may further involve determining a grid of points superimposed on an eye muscle boundary in frame 104. In some modalities, the segmentation model 220 can mask frame 104 to remove parts of the image that are not within the region of interest 132. In these cases, masking frame 104 can reduce the chance that points in the grid of points that are outside the determined region of interest will be counted erroneously.

[0095] In 310, the meat quality assessment module 214 (for example, the object quality determination module 218 or the feature prediction model 222) determines a prediction score for the first feature based on the determined region or area of ​​interest 132. The prediction score for the first feature may comprise a scalar value for the first feature corresponding to Petition 870250107908, dated 11 / 25 / 2025, page 37 / 58 30 / 45 first frame 104.

[0096] When operating in the second operating mode to assess the EMA, the meat quality assessment module 214 can determine the number of points within the identified region of interest 132. Since each point corresponds to one square centimeter, this determination provides a measure of the eye muscle area in cm2. In some modes, a different grid size may be used, such as a smaller grid or a larger grid. In modes where a smaller grid is used, the points may be less than one square centimeter apart. In modes where a larger grid is used, the points may be separated by more than one square centimeter. In these modes, a user can select the grid size via the user interface 120. In other modes, a user can select an area measure to which each point corresponds.

[0097] In 312, the meat quality assessment module 214 determines an assessment confidence rating for the first characteristic. The assessment confidence rating indicates whether sufficient information was determined from the video data to generate a quality assessment measure for the first characteristic.

[0098] In some modalities, the meat quality assessment module 214 determines the confidence rating of the assessment for the first characteristic, determining whether a sufficient number or a limit of suitable frames has been determined. If the total number of suitable frames determined does not exceed the limit (a minimum), the meat quality assessment module 214 determines that the confidence index of the assessment is not sufficient.

[0099] In some modalities, the meat quality assessment module 214 determines the confidence rating of the assessment for the first characteristic based on a function of the frame adequacy scores for the first characteristic of the determined adequate frames. For example, the frame adequacy scores for all non-excluded frames can be summed or added and compared to a threshold value (a Petition 870250107908, dated 11 / 25 / 2025, pp. 38 / 58 31 / 45 minimum). If the total of the adequacy scores of the table does not exceed the limit, the meat quality assessment module 214 determines that the confidence index of the assessment is not sufficient.

[0100] In some modalities, the meat quality assessment module 214 determines the confidence rating of the assessment for the first characteristic based on a function of the prediction scores for the first characteristic from the determined suitable frames. For example, the function of the prediction scores may be a sample standard deviation of the prediction score(s) for the determined suitable frame(s) for the characteristic. The sample standard deviation value may be compared to a threshold value (a maximum). If the sample standard deviation value exceeds the threshold, the meat quality assessment module 214 determines that the confidence rating of the assessment is insufficient.

[0101] In some embodiments, the meat quality assessment module 214 is configured to determine whether or not the frame is a suitable frame (e.g., 304 and 306) and to determine a prediction score for the frame (e.g., 308 and 310) in a substantially simultaneous or substantially parallel manner.

[0102] In some modes, the meat quality assessment module 214 determines only a prediction score for frames that were first determined as suitable frames. For example, in response to the determination of the frame as a suitable frame for predicting or assessing the first characteristic (e.g., in 304 and 306), the meat quality assessment module 214 determines a prediction score for the frame (e.g., in 308 and 310). In response to the determination of the frame as not being a suitable frame for predicting or assessing the first characteristic, the meat quality assessment module 214 may choose not to determine a prediction score for the frame (e.g., not run 308 and 310).

[0103] In step 314, in response to the determination that the confidence rating of the assessment for the first feature did not exceed Petition 870250107908, dated 11 / 25 / 2025, page 39 / 58 32 / 45 the confidence limit of the evaluation for the first characteristic, the meat quality evaluation module 214 runs 304 to 310 for a subsequent frame in the frame sequence, as, for example, can be received from the video capture device 112.

[0104] In step 316, in response to the determination that the confidence rating of the assessment for the first characteristic exceeded the confidence limit of the assessment for the first characteristic, the meat quality assessment module 214 determines a quality assessment measure for the first characteristic based on the prediction scores for the first characteristic from the determined suitable frames. Furthermore, in step 316, after the quality assessment measure for the first characteristic is determined, the meat quality assessment module 214 may send instructions to the processor 205 to terminate video capture via the video capture device 112.In modes, for example, where more than one characteristic is being evaluated, the meat quality evaluation module 214 can send instructions to the processor 205 to terminate video capture by the video capture device 112 after the completion of each or all characteristic quality evaluations. This allows video capture to be interrupted without the need for user input. The meat quality evaluation can be stored in memory 210. The meat quality evaluation module 214 can send instructions to the processor 205 to transmit the meat quality evaluation via the communication interface 226. In some modes, the image frames 104 stored in memory 210 can be transmitted via the communication interface 226. The communication interface 226 transmits data to an external source.Therefore, once sufficient frames for the accurate evaluation of a characteristic have been determined, no further frames are analyzed for that characteristic. This means that when a sufficient number of suitable frames have been acquired to allow the device to make an adequate assessment of meat quality, no further frames will be acquired and analyzed. This contrasts with a technique where a number... Petition 870250107908, dated 11 / 25 / 2025, pp. 40 / 58 A predetermined number of frames is acquired and evaluated, and in that case, it may be determined that the number of frames acquired is, in fact, insufficient to allow the evaluation to be carried out with sufficient accuracy, or that an excessive number of frames is acquired when, in fact, a subset of the frames would have been sufficient to allow the evaluation to be carried out with sufficient accuracy.

[0105] In some modalities, the meat quality assessment module 214 determines a quality assessment measure for the first characteristic using an outlier exclusion procedure and averaging as follows:

[0106] The second and first percentile values, Pb and Pa, respectively, of the prediction scores are calculated (where Pb ^ Pa). For example, the first and second percentile values ​​may be the 25th and 75th percentiles (first and third quartiles), respectively. In some modalities, the second and first percentile values, Pb and Pa, may be set to 20 and 80, respectively, or 10 and 90, respectively, or 40 and 60, respectively. However, it is important to note that any suitable value may be used.

[0107] Then, the interquartile range IQR = Pa - Pb is calculated.

[0108] Next, a lower limit Bmin is calculated by multiplying the interquartile range by a toliQR tolerance parameter (configurable, with a default value of 1.5) and subtracting it from the value of the 25th percentile: Bmán= PB~ (tolIQRx IQR)

[0109] Similarly, an upper limit Bmax is calculated as: BmáX= PA+ (t°lIQRx IQR)

[0110] Finally, the arithmetic mean is obtained from all prediction values ​​of individual frames that fall within the interval [Bmin, Bmax] to produce the quality assessment measure of the first feature.

[0111] In some modalities, when determining the measure Petition 870250107908, dated 11 / 25 / 2025, pp. 41 / 58 In the 34 / 45 quality assessment, the 214 meat quality assessment module can determine and ignore or disregard prediction scores derived from the “worst” frame drop percentage based on frame suitability ratings (e.g., for each of the brightness, sharpness, and segmentation sizes). In other words, in these modalities, these prediction scores are not used to determine the characteristic quality assessment measure.

[0112] In some modalities, the quality assessment measure may be determined by averaging prediction scores obtained without exclusion. In some modalities, the quality assessment measure may be determined by the median or geometric mean of the prediction scores (as opposed to the arithmetic mean). In some modalities, the quality assessment measure may be determined by a weighted average using the corresponding adequacy scores; this may allow frames 104 with higher adequacy scores to have more influence on the resulting combination. In some modalities, outliers may be excluded from the quality assessment measure by determining what multiple of the standard deviation (of the sample) each prediction is from the mean and comparing them to an adequate threshold.In some methods, the quality assessment measure can be determined by grouping the forecasts and selecting a specific group according to a criterion. The criteria can be the largest cluster or a number of the largest clusters, such as the n largest clusters.

[0113] After the meat quality assessment module 214 has determined the quality assessment measure, it can output the result to the user via the user interface 120. For example, the user interface module 212 can cause the quality assessment measure of one or more characteristics to be displayed on the screen 130 and / or can store the quality assessment measure(s) in memory 210.

[0114] When all the animal's characteristics reach this point, video capture from video capture device 112 may be terminated by processor 205, and the user may be prompted to switch to the Petition 870250107908, dated 11 / 25 / 2025, pp. 42 / 58 35 / 45 next casing for analysis via an alert in the user interface 120.

[0115] In some modes, an option may be included to define a maximum number of Nf frames for analysis (which applies to all features and includes all frames, including those that were excluded from the processing of one or more features). If the total number of frames analyzed exceeds this number, the first frame in the received sequence (the oldest frame) will be excluded from future calculations and only the most recent Nf frames will be used for features for which a quality assessment measure has not yet been produced.

[0116] In some modalities, the meat quality assessment module 214 can evaluate the frames for a second characteristic and / or additional characteristics according to method 300. The second characteristic and / or additional characteristics are different from the first characteristic and from each other. This evaluation using method 300 can be performed almost simultaneously with the evaluation of the first characteristic using method 300.

[0117] As mentioned above, in some embodiments, the video capture device 112 may comprise a 3D camera configured to determine depth information, which can be used to evaluate the EMA characteristic. However, it is important to note that depth information can also be used to evaluate marbling characteristics, rib eye area, rib fat thickness, meat color, fat color, marbling fineness, or intramuscular fat ratio. The processor 205 can receive instructions from the user interface 120 to start the 3D camera to record 3D video data. The 3D video data may comprise video image frames 104 obtained in more than one orientation. The 3D video image data comprising video image frames 104 obtained from more than one orientation can be processed by the 3D camera interface module 213.The 3D camera interface module 213 can determine a 3D point cloud based on captured 3D video data. The 3D camera interface module 213 can cross-reference the frame of... Petition 870250107908, dated 11 / 25 / 2025, pp. 43 / 58 36 / 45 image 104 to produce a 3D model of an object in the camera's field of view. This may comprise a meat product 140. In some embodiments, this may comprise a portion of a meat product 140. The 3D camera interface module 213 may communicate with the meat quality assessment module 214. The 3D camera interface module may send the 3D point cloud to the object determination module 216. The object determination module 216 may receive the point cloud from the 3D camera interface module 213 and isolate the area of ​​the point cloud corresponding to the region of interest 132. In some embodiments, another target area of ​​the meat product 140 may be identified in the point cloud. This may comprise the region of interest 132. The 3D camera interface module 213 may send the 3D point cloud to the meat quality assessment module 214.The meat quality assessment module 214 can determine a plane, fitted to the points of the 3D point cloud. The meat quality assessment module 214 can project the points of the 3D point cloud onto the plane. This can generate a plurality of projected planar points that can be processed later. The area of ​​the plane can be determined by the meat quality assessment module 214 by positioning the projected points on the plane. The determined area of ​​the region of interest 132 can correspond to the area of ​​the eye muscle.

[0118] The projected planar points can then be rasterized into a high-resolution image in which each pixel cell of the image corresponds to a square region of the plane of known real-world size, i.e., an area of ​​the real world. For example, the real-world area can be determined by the meat quality assessment module 214 based on video data. In some embodiments, the pixel cell size may be between 0.5 mm2 and 0.01 mm2. In some embodiments, the pixel cell size may be 0.25 mm2. The square region of the plane of known real-world size may contain “holes,” i.e., pixel cells within the region of interest that do not contain a corresponding point in the planar projection of the point cloud. The high-resolution image can then be converted to binary, in which each pixel has a Petition 870250107908, dated 11 / 25 / 2025, pp. 44 / 58 37 / 45 value of 1 or 0, depending on whether or not it contains a point in the planar projection. The resulting binary image can identify the area of ​​the region of interest.

[0119] The meat quality assessment module 214 can then be configured to fill in the holes in the image using a morphological closure operation. For example, a dilation operation followed by an erosion operation can be applied, using the same structuring element for both operations. The morphological operation can use a core of appropriate size. The morphological closure operation can be effective in filling small holes in the image while preserving the shape and size of any large holes and objects in the image. The meat quality assessment module 214 can then use contour detection or contour recognition to obtain the edge of the region of interest in the filled binary image. In some modes, the detection of a single contour is confirmed and / or verified.In other words, when the filled region of interest is confirmed as contiguous in the binary image, several possible failure modes are filtered out. In response to the identification of a single contour, the known correspondence of the pixels with the real-world area is used to take the area within the contour and convert it into an estimate of the area of ​​the real-world region of interest.

[0120] In this way, the meat quality assessment module 214 can determine the size of the eye muscle area (EMA) based on the determined area of ​​the region of interest 132. Since the cut surface of a meat product 140 can be considered approximately flat, errors due to noise or depth measurement can be corrected based on the approximately flat surface.

[0121] In some embodiments, the video capture device 112 may comprise a 3D camera configured to determine depth information, which may be used to assess the rib fat thickness characteristic. The meat quality assessment module 214 may be configured to measure the rib fat thickness characteristic. Petition 870250107908, dated 11 / 25 / 2025, pages 45 / 58 38 / 45 Similar to the process of isolating the region of interest, the 3D camera interface module 213 can communicate with the meat quality assessment module 214 and can send a determined 3D point cloud to the object determination module 216. The object determination module 216 can receive the point cloud from the 3D camera interface module 213 and isolate the areas of the point clouds corresponding to the eye muscle (EMA) and the area of ​​the point cloud corresponding to the rib fat. In some embodiments, the EMA and rib fat points are isolated in the 3D point cloud by referencing a neural network-derived segmentation of the color image component of a depth map camera frame from the video data, where each pixel corresponds to a point in the 3D point cloud, since the cross-reference is a depth map image.In some techniques, isolation of the rib fat and rib muscle can be performed sequentially or in parallel.

[0122] The 3D camera interface module 213 can send the 3D point cloud to the meat quality assessment module 214. The meat quality assessment module 214 can determine a plane, fitted to the points of the 3D point cloud. The meat quality assessment module 214 can project the points of the 3D point cloud onto the plane. This can generate a plurality of projected planar points that can be processed further. The area of ​​the plane can be determined by the meat quality assessment module 214 by positioning the projected points on the plane. In this case, the determined area corresponds to the EMA and rib fat. The plurality of projected planar points represents a 2D point cloud with EMA and rib fat points. Because they were isolated separately from each other, the projected points for the EMA are distinct from the projected points for the rib fat.

[0123] The first principal component (PC1) of the EMA is identified and the 2D point cloud is aligned so that PC1 is vertical. In some embodiments, it is assumed that the camera is held in the correct position and that filming occurs from the bottom of the 180-degree EMA. This may be Petition 870250107908, dated 11 / 25 / 2025, pages 46 / 58 39 / 45 is used to choose which of the two possible "upward paths" should be used.

[0124] The projected planar points can then be rasterized into an image. In some embodiments, a coarser grid of pixel cells is used than that used in the area calculation process of the region of interest. For example, a 1 mm² pixel cell is used. The size of the pixel cells used can be chosen to minimize or reduce the number of holes. In some embodiments, the coarseness of the pixel cell grid can be selected so that it is coarse enough that no holes are expected. That is, each pixel can contain or encompass at least one point of the planar projection. This provides a lower resolution 2D image (color, non-binary) with the EMA and rib fat, where it is known which pixels correspond to the EMA and which pixels correspond to the rib fat, and where the correspondence between the pixels and the real-world geometry is known.

[0125] In some modalities, the method for measuring rib fat thickness may vary depending on the jurisdiction in which the method is being performed.

[0126] To measure rib fat thickness, the y-axis line is identified at an appropriate percentage of the distance between the bottom and top of the EMA that meets the rib fat thickness measurement requirements. The top of the EMA can be aligned vertically along its first principal component, as described previously. In some embodiments, rib fat thickness can be measured approximately three-quarters of the way from the EMA to the top. In this case, the y-axis line value may be approximately 75%. In some embodiments, it may be in the range of 70% to 85%. In some embodiments, the y-axis line value may be 79%. However, it is important to note that this value can be determined empirically and may vary according to video data and meat product.

[0127] The meat quality assessment module 214 can then be configured to measure rib fat thickness along Petition 870250107908, dated 11 / 25 / 2025, pp. 47 / 58 40 / 45 of the y-axis line, a predetermined number of lines around the y-axis line. For example, the lines between 74% and 83% of the way to the EMA. In some embodiments, the surrounding lines can be determined based on a pre-selected threshold or buffer. The number of adjacent lines below the y-axis line value may differ from the number of adjacent lines above the y-axis line value. In some embodiments, thickness can be measured in pixels. In some embodiments, the number of surrounding lines along which rib fat thickness is measured can be determined empirically and may vary depending on video data and meat product.

[0128] Pixel thicknesses can be converted back into real-world thickness estimates for rib fat. The out-of-curve point exclusion and averaging procedure described in this document can be used to aggregate the lines into an overall estimate for that frame. In some embodiments, the frames themselves are aggregated using a similar procedure. In some embodiments, frames may be aggregated according to suitability filtering, for example, where only frames determined to meet a specific suitability classification are aggregated. In some embodiments, when the difference between the largest and smallest thickness estimate per line is too large in a given frame, the frame may be rejected for rib fat thickness measurement.In some methods, when many of the lines within the percentage range being used do not contain any rib fat points, the frames may be rejected for rib fat thickness measurement.

[0129] It should be considered that procedures similar to those described in this document regarding rib eye area and rib fat thickness can also be used to determine measures related to other characteristics, such as marbling, rib eye area, meat color, fat color, marbling fineness, or intramuscular fat ratio. In some embodiments, rib eye area, rib fat thickness, meat color, fat color, marbling fineness, or intramuscular fat ratio Petition 870250107908, dated 11 / 25 / 2025, pp. 48 / 58 41 / 45 Intramuscular fat can be assessed by analyzing video data from 2D or 3D cameras. In some modalities, EMA, marbling, rib eye area, rib fat thickness, meat color, fat color, marbling fineness, or intramuscular fat ratio can be assessed using depth and / or non-depth analysis modes.

[0130] Figures 4a and 4b are examples of screenshots of user interface displays of the analysis device of Figure 2, according to some embodiments. Figure 4a shows the display 130 with a video data frame 402 representing a portion of a carcass to be evaluated for quality. The video data 402 may be a live output from the video capture device 112. The video data 402 may be represented with segments 410, 412 that represent the result of the segmentation model 220. Segments 410 may represent a specific segment corresponding to a portion of fat in the carcass. Segments 412 may represent a specific meat product 140 in the carcass. In the embodiment of Figure 4a and Figure 4b, the specific meat product 140 is a rib cut.The segments to be determined may be shown alongside the video data 402, as shown in 404 on screen 130, 406 may comprise at least one progress indicator of the confidence limit for the animal characteristics being evaluated. 406 may be a progress bar 406 that shows a graphical representation of the progress of the number of suitable frames determined for each characteristic being evaluated. In some modalities, 406 is a percentage indicator. In some modalities, 406 is a numerical indicator. In some modalities, 406 is a word indicator.

[0131] In 408, the display 130 can represent operational icons that allow the display 130 to provide the functions of the user interface 120, such as on-screen buttons to control the operation of the analysis device 110 or to provide the ability to control the operation of the meat quality assessment module 214. For example, in Figure 4a, the analysis device 120 may be in the middle of executing method 300. In this way, the operational buttons in Petition 870250107908, dated 11 / 25 / 2025, pp. 49 / 58 42 / 45 408 allow the display 130 to act as a user interface 120 to allow a user to interrupt the meat evaluation operation. In the mode of Figure 4b, which shows the same carcass portion after the meat evaluation is complete, the operational buttons on 408 allow the method 300 to be restarted on a new carcass, a video to be marked for later evaluation, or the evaluation modes to be changed (e.g., between EMA and normal operation). Thus, as Figure 4b shows a completed meat quality evaluation, 406 shows the completed evaluation for the respective animal characteristics. In the mode of Figure 4b, these are: marbling (Australian and MSA grades), meat color, fat color, and intramuscular fat percentage.By displaying the carcass part being evaluated 402, showing the segmentation portions 412, 410, having context-sensitive operating buttons in 408 and the progress / result of the meat quality evaluation in 406, 418, a user can easily and effectively operate an analysis device 110 to perform the method 300. These operations and the display of the user interface features described in Figures 4a and 4b are provided by the executable code stored in the user interface module 212.

[0132] In embodiments of Figures 4a and 4b, the user interface is optimized to work with a touch screen, such as the display of a smartphone. However, in other embodiments, the display 130 may be a simple display screen without touch screen functionality. In these embodiments, the operating buttons in 408 and 416 may be physical buttons as part of the analysis device 110. Alternatively, more than one display 130 may be provided to display different information – such as the progress indicators of 406 or the result of the meat quality assessment in 418. In some embodiments, this information may be presented as part of the user interface 120 by lights, LCD screens or other visual indicators to provide the operator with an indication of the result of method 300 for certain animal characteristics.

[0133] In some modes, the 3D point clouds generated by the 3D depth camera can be used to obtain a frame or a set of frames, captured from any angle of the cutting surface of a Petition 870250107908, dated 11 / 25 / 2025, pages 50 / 58 43 / 45 meat product, to generate a high-resolution synthetic visualization of the cut surface. For example, as if it had been filmed from top to bottom, perpendicularly, or in an aerial view. In some embodiments, the angle of the cut surface may be within a reasonable range. In some embodiments, 3D point clouds can also be used to generate high-resolution synthetic visualizations of EMA, or measurements of EMA and rib fat thickness, rib eye area, meat color, fat color, marbling fineness, or intramuscular fat ratio. In some embodiments, video data captured by a 2D camera (or one or more frames of the video data) can be used to create a high-resolution synthetic image of a top-down view.In some embodiments, a sequence of high-resolution synthetic images of a top-down view of the cut surface of the meat product can be generated using video data obtained by the 2D camera. In some embodiments, a neural network and / or machine learning approach can be used to generate the images, for example, by interpolating, extrapolating, or generating a synthetic image of the cut surface based on the obtained video data. In some embodiments, the high-resolution synthetic image (or a sequence of high-resolution synthetic images) can be created using a combination of video data obtained from the 3D and 2D cameras. That is, the synthetic image of the cut surface can be created using depth video data and normal video data.

[0134] The high-resolution synthetic image can be generated directly on the analysis device 110, which includes the video capture device 112. In some embodiments, the synthetic image can be generated outside the analysis device 110, for example, on a server, where the point cloud data is transmitted to an external server and then configured to receive the synthetic image generated from the server. The synthetic image can be generated in real time during data capture or after data capture has been performed.

[0135] The generation of the synthetic image from the data of Petition 870250107908, dated 11 / 25 / 2025, pages 51 / 58 44 / 45 point cloud generation may involve obtaining a low-resolution rasterized image as a result of determining rib fat thickness measurements, and applying upscaling techniques to take it from the low resolution imposed by the source pixel grid to a high resolution. In some embodiments, the application of upscaling techniques may include the use of a generative ML model. Synthetic image generation from point cloud data may involve obtaining a 2D planar projection point cloud and applying one or more interpolation techniques before converting the point cloud into a high-resolution rasterized image using a finer pixel grid. In some embodiments, the interpolation techniques may be traditional or machine learning-based techniques.In some embodiments, the generation of the synthetic image may include obtaining a high-resolution rasterized image, as a result of determining the area of ​​the region of interest, as described in this document, which was obtained using a finer pixel grid, and “filling” the holes with a painting technique. In some embodiments, the painting technique may use traditional methods or machine learning, for example, a diffusion machine learning model.

[0136] In some embodiments, the generation of the synthetic image may include obtaining the 3D point cloud (or isolated EMA points or EMA and rib fat thickness, or any other isolated part or feature points within the point cloud) and using a direct means to obtain a high-resolution rasterized image. For example, a neural network may be used to obtain a high-resolution rasterized image of the 3D point cloud. In some embodiments, the generation of a synthetic image may include obtaining the 2D planar projection point cloud (or isolated EMA points or EMA and rib fat thickness, or any other isolated part or feature points within the point cloud) and using a direct means to obtain a high-resolution rasterized image of the 2D point cloud. For example, a neural network may be used to obtain a high-resolution rasterized image.In some applications, the generation of the synthetic image may include... Petition 870250107908, dated 11 / 25 / 2025, pp. 52-58 45 / 45 using any of the techniques described in this document, taking into account more than one frame at a time to produce a single resulting high-resolution synthetic image.

[0137] In some embodiments, all frames of the video data can be used to produce the single resulting high-resolution image. The use of a plurality of frames can take advantage of information about the temporal sequence of the frames used in the process. In some embodiments, a plurality of frames can be used without using frame ordering or temporal information. The generated high-resolution synthetic image may be in the form of a sequence of frames and / or a video of high-resolution images, each derived from a frame. In some embodiments, the generation of the synthetic image may include the use of a combination of two or more techniques described in this document.

[0138] The generated high-resolution synthetic image can be used for further processing. For example, for additional feature evaluation / analysis. This additional processing can be performed on the analysis device 110 or externally to the analysis device 110. The generated high-resolution synthetic image can be used for visual display. For example, on the analysis device 110 or externally to the analysis device 110, such as via some hardware accessory equipped with a screen or via an application, such as an online web portal. In some embodiments, the synthetic image can be displayed in real time during data capture or after data capture.

[0139] Those skilled in the art will realize that numerous variations and / or modifications can be made to the modalities described above, without departing from the broad general scope of this disclosure. The present modalities should, consequently, be considered in all respects as illustrative and not restrictive. Petition 870250107908, dated 11 / 25 / 2025, pages 53 / 58

Claims

1 / 6 CLAIMS 1. A method characterized in that it comprises: determining video data comprising a sequence of frames, wherein at least some of the frames represent a portion of a carcass to be evaluated for quality; for a first frame in the sequence of frames: a) determining a frame suitability score for evaluating a first feature; b) in response to the fact that the frame suitability score is greater than the frame suitability threshold score for the first feature, determining the frame as a suitable frame for evaluating the first feature; c) determining, by means of a segmentation model, a region of interest in the frame for evaluating the first feature; d) determining, by means of a first feature prediction model, a prediction score for the first feature based on the determined region of interest;(ee) determine a confidence rating of the assessment for the first feature; in response to the determination that the confidence rating of the assessment for the first feature did not exceed the confidence limit of the assessment for the first feature, perform steps (a) and (e) for a subsequent frame in sequence; and in response to the determination that the confidence rating of the assessment for the first feature exceeded the confidence limit of the assessment for the first feature, determine a quality assessment measure of the first feature based on the prediction score for the first feature of the determined suitable frames.

2. Method, according to claim 1, characterized in that determining the confidence rating of the assessment for the first feature comprises: determining one or more of: (i) a number of suitable frames determined for the first feature; and (ii) a function of the frame suitability scores for the first feature of the suitable frames determined.

3. Method, according to claim 1 or 2, characterized in that determining the confidence rating of the assessment for the first feature comprises: determining a function of the prediction scores for the first feature of the determined suitable frames.

4. Method, according to claim 3, characterized in that determining a function of the prediction scores for the first characteristic of the determined suitable frames comprises determining an arithmetic mean obtained in the interval: [Bmin, Bmax] where Bmin is defined by the function: and where Bmax is defined by the function: where Pa is a prediction value of a first percentile of the prediction scores of the first characteristic; where Pb is a prediction value of a second percentile of the prediction scores of the first characteristic; where toi^ is a tolerance parameter; and where IQR is an interquartile range defined by IQR = Pa-Pb.

5. Method according to claim 4, characterized in that it further comprises: excluding the lowest pdrop percentage of prediction scores Petition 870250087453, dated 09 / 26 / 2025, page 119 / 125 3 / 6 for one or more of: (i) a measure of lack of brightness; (ii) a measure of sharpness; and (iii) a measure of segmentation size; wherein pdrop is a configurable parameter.

6. A method according to claim 4 or 5, characterized in that it further comprises: determining a frame number limit for the frame sequence; in response to the determination of the number of frames exceeding the frame number limit, discarding the frame adequacy score, the prediction score, and the confidence rating for the first frame determined.

7. A method, according to any one of claims 3 to 6, characterized in that the first percentile is a percentile lower than the second percentile.

8. Method according to claim 1 or 2, characterized in that it further comprises: in response to the determination of the frame as a suitable frame for evaluating the first characteristic, performing steps c) and d); and in response to the determination of the frame as not being a suitable frame for evaluating the first characteristic, omitting steps c) and d).

9. A method, according to any of the claims, characterized in that determining the video data comprises receiving a video stream.

10. Method, according to any of the claims, characterized in that the frame adequacy score is based on one or more of: (i) a measure of lack of brightness, (ii) a measure of sharpness, (iii) a measure of segmentation size; (iv) a measure of segmentation margin of the region of interest; (v) a measure of roundness of the segmentation of the region of interest.

11. Method, according to any of the claims, Petition 870250087453, dated 09 / 26 / 2025, p. 120 / 125 4 / 6 characterized in that determining the frame adequacy score comprises determining the frame adequacy score according to the following equation: St = max (vracr + b^, 0) where st is the adequacy score; l> is a classification type r value for the first frame; ar is a classification type r scaling coefficient for feature f; and br is a classification type r compensation coefficient for feature t.

12. A method, according to any of the preceding claims, characterized in that it further comprises: issuing a quality assessment measure for the first feature of a user interface.

13. A method, according to any of the preceding claims, characterized in that the first characteristic comprises any of the following: eye muscle area, marbling, fineness of marbling, rib area, rib fat thickness, intramuscular fat, fat color, and meat color.

14. Method, according to any of the preceding claims, characterized in that it further comprises: for the first frame in the sequence of frames: f) determining a frame adequacy score for evaluating a second feature, wherein the second feature is different from the first; g) in response to the fact that the frame adequacy score is greater than the frame adequacy threshold score for the second feature, determining the frame as an adequate frame for evaluating the second feature; Petition 870250087453, dated 09 / 26 / 2025, p. 121 / 125 5 / 6 h) determining, by means of a second segmentation model, the region of the area of ​​interest in the frame for evaluating a second feature; i) determining, by a second feature prediction model, a prediction score for the second feature based on the determined region of interest;(e) determine a confidence rating of the assessment for the second characteristic; in response to the determination that the confidence rating of the assessment for the second characteristic did not exceed the confidence limit of the assessment for the second characteristic, perform steps (a) and (e) for a subsequent frame in sequence; and in response to the determination that the confidence rating of the assessment for the second characteristic exceeded the confidence limit of the assessment for the second characteristic, determine a quality assessment measure of the second characteristic based on the prediction scores for the second characteristic of the appropriate frames determined.

15. Method, according to claim 14, characterized in that the second characteristic is different from the first characteristic and comprises any of the following: eye muscle area, marbling, fineness of marbling, rib eye area, rib fat thickness, intramuscular fat, fat color and meat color.

16. A method, according to any of the claims, characterized in that the frame adequacy score is based on one or more frame adequacy ratings, each frame adequacy rating indicating the adequacy of a frame with respect to a specific measure, and the method further comprises, for each of the frame adequacy ratings: comparing the frame adequacy rating with a respective specific rating threshold of the measure; and Petition 870250087453, dated 09 / 26 / 2025, p. 122 / 125 6 / 6 responsive to the determination that the frame adequacy rating does not meet the threshold, determining the frame as inadequate and excluding it from further processing.

17. Meat evaluation system characterized in that it comprises: at least one processor; memory accessible to at least one processor; and comprising computer executable instructions which, when executed by at least one processor, cause the system to perform the method as defined in any one of claims 1 to 16.

18. System according to claim 17, characterized in that it further comprises: a user interface configured to display the quality assessment measure determined for the first characteristic.

19. System according to claim 17 or 18, characterized in that it further comprises a video capture device for capturing video data, controlled by at least one processor.

20. Non-transient machine-readable media characterized in that it stores instructions which, when executed by one or more processors, cause an electronic device to perform the method as defined in any one of claims 1 to 16. Petition 870250087453, dated 09 / 26 / 2025, pp. 123 / 125

Images