246 results about "Object detector" patented technology

The invention discloses a self-prompting semantic segmentation method and device based on SAM and a storage medium, and belongs to the technical field of image segmentation. The method comprises the following steps: acquiring a target image containing a target object; inputting the target image into a semantic segmentation model based on SAM to obtain a segmentation result of the target image; the segmentation result comprises the position and the type of the defect of the target object; wherein the semantic segmentation model comprises an image encoder, a target detector, a prompt encoder and a mask decoder; the target detector is used for positioning the position area of the defect of the target object on the target image according to the image features output by the image encoder, and outputting a defect detection frame; the prompt encoder is used for encoding the defect detection frame and generating position prompt characteristics; and the mask decoder is used for fusing the image features and the position prompt features and generating a segmentation result of the target image. According to the invention, end-to-end industrial defect detection can be realized, and the efficiency of industrial defect detection is improved.

The invention belongs to the technical field of ship collision avoidance, discloses a ship collision avoidance early warning method and system based on vision and AIS data collaborative perception, and firstly provides an enhanced YOLO (E-YOLO) target detector which can detect a ship in a complex environment in a high-precision and real-time manner and perform ship tracking and visual trajectory extraction in combination with a DeepSORT algorithm. Secondly, a homography coordinate conversion method based on multiple frames of static images is provided, conversion between a pixel coordinate system and a world coordinate system can be realized without accurate camera calibration, and the fusion precision of visual data and AIS data is improved. AIS and visual information are matched through a K-D tree algorithm, AIS data are projected to visual coordinates, and accurate matching of heterogeneous data is achieved. The system is provided with ship illegal behavior judgment and collision early warning rules, behaviors such as overspeed and abnormal AIS closing can be recognized, and early warning is triggered. Experiments prove that the method provided by the invention is strong in adaptability and high in robustness, and can effectively improve the ship monitoring and traffic management level of the bridge water area.

Techniques for determining a presence of an object, especially an object such as animal or debris, in a path of a vehicle, are discussed herein. For example, sensors of various modalities, which may include multispectral sensors, may capture data representing an environment the vehicle is traversing. In examples, one or more trained machine learned (ML) models, operating on a vehicle computing system, may detect and / or classify objects in the environment, based on input data of one or more modalities or spectral bands. The ML models may be pre-trained using training data including real sensor data, synthetic data, and / or augmented data, along with auto-generated annotations. In some examples, hyperspectral data may be used to identify materials associated with detected objects. A confidence score associated with the detection of the object may also be computed. The vehicle may be controlled based on detection of the object and its classification.

The invention relates to a computer-implemented method for automatically detecting anatomical structures (3) in a medical image (1) of a subject, the method comprising applying an object detector function (4) to the medical image, wherein the object detector function performs the steps of: (A) applying a first neural network (40) to the medical image, wherein the first neural network is trained to detect a first plurality of classes of larger-sized anatomical structures (3a), thereby generating as output the coordinates of at least one first bounding box (51) and the confidence score of it containing a larger-sized anatomical structure; (B) cropping (42) the medical image to the first bounding box, thereby generating a cropped image (11) containing the image content within the first bounding box (51); and (C) applying a second neural network (44) to the cropped medical image, wherein the second neural network is trained to detect at least one second class of smaller-sized anatomical structures (3b), thereby generating as output the coordinates of at least one second bounding box (54) and the confidence score of it containing a smaller-sized anatomical structure.

An input device determines presence of an actual user, instead of an artifact, by using multi-wavelength reflectance spectroscopy. Light sources are operated to illuminate an object with different colors of light at different times. A detector determines, at those different times, intensity data indicative of intensity light of these different colors as reflected from the object. The intensity data is processed to determine whether the object is part of a user or is an artifact. For example, if the object is deemed to be a user, biometric input may be acquired. The biometric input may then be processed to identify the user. The input device may be used at various locations, such as at an entry portal, point of sale, and so forth.

The invention relates to the technical field of target detection, and provides a track operator safety wearing detection method, device and equipment and a storage medium, and the method comprises the steps: collecting a video image of a track operation construction site, and inputting the video image into a pre-trained safety wearing detection model, detecting the safety wear of each operator in the video image to obtain a detection result of the safety wear of each operator; wherein the safe wearing detection model is constructed by taking a memory efficient visual converter with a cascade grouping attention mechanism as a backbone network of a real-time target detector model and taking an ADown downsampling layer as a downsampling layer of the real-time target detector model. Through the constructed safe wearing detection model, the detection efficiency and the detection effect of safe wearing of the operating personnel are improved, the requirements of real-time performance and accuracy of safe wearing condition detection of the operating personnel in a rail transit scene can be met, and the workload of safety patrol of management personnel is reduced.

In various examples, detected object data representative of locations of detected objects in a field of view may be determined. One or more clusters of the detected objects may be generated based at least in part on the locations and features of the cluster may be determined for use as inputs to a machine learning model(s). A confidence score, computed by the machine learning model(s) based at least in part on the inputs, may be received, where the confidence score may be representative of a probability that the cluster corresponds to an object depicted at least partially in the field of view. Further examples provide approaches for determining ground truth data for training object detectors, such as for determining coverage values for ground truth objects using associated shapes, and for determining soft coverage values for ground truth objects.

Techniques are discussed herein for generating three-dimensional (3D) representations of an environment based on two-dimensional (2D) image data, and using the 3D representations to perform 3D object detection and other 3D analyses of the environment. 2D image data may be received, along with depth estimation data associated with the 2D image data. Using the 2D image data and associated depth data, an image-based object detector may generate 3D representations, including point clouds and / or 3D pixel grids, for the 2D image or particular regions of interest. In some examples, a 3D point cloud may be generated by projecting pixels from the 2D image into 3D space followed by a trained 3D convolutional neural network (CNN) performing object detection. Additionally or alternatively, a top-down view of a 3D pixel grid representation may be used to perform object detection using 2D convolutions.

A computer-implemented method for training a machine learning, ML, model to perform object detection, the method comprising: obtaining a first training dataset comprising a plurality of unlabelled images, each unlabelled image containing at least one object; analysing the first training dataset by using an object detector module of the ML model; forming a second training dataset using the unlabelled images of the first training dataset and their corresponding extracted bounding boxes and pseudo-labels; and training the object detector module, using the second training dataset, to output bounding boxes and pseudo-labels for input pseudo-labelled images.

The invention belongs to the field of computer vision and artificial intelligence security, and discloses a self-adaptive frequency domain adversarial training method and device for a target detector, the target detector comprises a repair module and a pedestrian detector which are connected in series, and the input of the repair module is connected with the output of a patch detector; the self-adaptive frequency domain adversarial training method comprises the following steps: losses in joint training comprise standard target detection losses, repair consistency losses on a frequency domain based on a frequency domain image corresponding to a training image and a clean image, and repair dependence losses based on a detected average precision mean value; according to the invention, the end-to-end joint training is carried out through the restoration module and the subsequent pedestrian detector, and the optimization target of the restoration module is directly aligned with the improvement of the detection robustness, so that the confrontation disturbance is eliminated as far as possible, and meanwhile, the key semantic information of the detection task is reserved to the maximum extent. The separation of the performance of the repair module and the pedestrian detector is avoided, and the detection robustness is improved.

Systems and methods for a multi-camera object detector having two or more backbone models. In particular, systems and methods are provided for including two or more backbone machine learning models, with one backbone optimized for speed and the other backbone optimized for precision. In particular, the first backbone can be highly accurate but slower than the second backbone and with a higher computer resource usage. The second backbone can be fast and efficient but have lower accuracy for object detection. In some examples, the second backbone can use fewer images and / or lower resolution images. The determination of which backbone to use can be based on fixed rules, or it can be determined based on another machine learning component. The outputs from the first and second backbones for each camera can be combined together into a unified representation, such as a bird's eye view (BEV) space.

Systems, methods, and storage media for manifesting a virtual object in a virtual environment are disclosed. Exemplary embodiments may: receive, at a first physical object detector, a first signal, from a first physical object-associated element in a first physical environment; identify, at a first value identification module, based on the first signal, a first value associated with the first signal; identify, at a first virtual object identification module, based on the first value, a first virtual object; and manifest, at a first virtual environment output device, a first manifestation of the first virtual object in a first manifestation of the first virtual environment.

A fixed object detector includes: a capturing device; a fixed-object area information obtainer (a dirt area information obtainer) that searches for a fixed object candidate that is a candidate for a fixed object appearing in a video captured by the capturing device, and obtains a fixed-object area information item indicating coordinates of the fixed object candidate in the video, the fixed object having a fixed relative position with respect to the capturing device; a storage that stores therein the fixed-object area information item obtained; and an observer that determines that the coordinates indicated by the fixed-object area information item to be coordinates of the fixed object in the video and outputs a determination result, the coordinates indicated by the fixed-object area information item matching coordinates indicated by a past fixed-object area information item stored in the storage in past.

A method for automated navigational marker detection and association in maritime applications includes providing a camera system equipped with one or more red-green-blue (RGB) cameras for capturing visual data of a surrounding maritime environment in real-time, and then providing a computing unit comprising a neural network-based object detector, a projection mechanism module, and a GPS mapping and chart data integration module. Next, providing a database comprising pre-existing chart data of navigational markers. Next, capturing visual data of the surrounding maritime environment using the camera system. Next, processing the visual data by the computing unit using the neural network-based object detector to identify navigational markers. Next, projecting pixel positions of detected navigational markers into a three-dimensional (3D) coordinate system. Next, integrating the projected positions of the detected navigational markers into a navigational map, and then cross-referencing the projected positions of the detected navigational markers with pre-existing chart data of navigational markers for the same location to enhance navigational accuracy and reliability.

A method for generating training data for an object detector. The method includes receiving a plurality of optical images of a scene, each camera showing the scene from a respective viewing direction of a plurality of different viewing directions, receiving a plurality of sensor data elements, each sensor data element including sensor data other than optical image data of the scene from a respective sensing direction of a plurality of different sensing directions, training a first neural radiance field using the plurality of optical images to generate, for each 3D point of the scene, a respective value of a predetermined feature, training a second neural radiance field using the plurality of sensor data elements to generate, for each 3D point of the scene, a respective sensor data value and generating training data elements for the object detector using the first and the second neural radiance field.

The invention belongs to the field of dynamic scene graph generation in scene graph generation sub-problems, and particularly relates to a dynamic scene graph generation method based on visual semantic reasoning, which comprises the following steps of: acquiring a video, inputting the video into a trained dynamic scene graph generation model, obtaining objects and relationship categories among the objects, and generating a dynamic scene graph; constructing a scene graph structure of the video according to the objects and the relationship categories among the objects; the dynamic scene graph generation model comprises a pre-trained object detector, a time relation encoding module, a relation feature extraction module, a spatial knowledge encoder, a time-space relation decoding module and a classification module; according to the method, visual feature differences and semantic feature differences between adjacent video frames are calculated, loss calculation is carried out, and a direction is provided for correct prediction of a model; according to the method, object space information and time dependence are fused through a space encoder, correct relation labels are screened through priori knowledge, and the accuracy of relation prediction is improved.

An image processing apparatus includes a video obtainer that obtains a video captured with a camera, a human detector that performs human detection with the obtained video, a moving object detector that performs moving object detection with the obtained video, a human candidate identifier that identifies, as an image of a human candidate area, an image of an area detected through human detection by the human detector based on a degree of matching between the image of the area detected through human detection by the human detector and an image of an area detected through moving object detection by the moving object detector, and a determiner that determines whether the identified image of the human candidate area is an image of a human based on a degree of matching between the image of the human candidate area and a reference image of an object erroneously detected as a human.

Unlike existing methods, using LIDAR and 2D-cameras for detection and limit in providing robust tracking of objects, the method disclosed addresses the technical challenge by utilizing 3D-LIDAR points clouds, 2D-camera images and additionally 2D-BEV from 3D LIDAR point clouds to provide, robust, seamless, 360 tracking of objects. The method independently detects, and tracks objects captured by each of the LIDAR, 2D-camera set up mounted on an autonomous vehicle. Detection and tracking is performed on 2D-Camera and 2D-BEV and fused with 3D-Lidar tracker, Additionally, non-occluded area of an object in 2D-bounding boxes, is identified by superimposing panoptic segmentation output with 2D-BB, that helps to eliminate the lidar points falling on the irrelevant objects and provides accurate real world position of the detected objects in the point cloud. An enhanced 2D object detector based on a customized NN architecture employing MISH activation function is also introduced.

Example solutions for ranking object detection results generate or receive a plurality of segmentation masks each corresponding to one or more images. Each segmentation mask of each plurality of segmentation masks is generated using a different object detector or setting options. A quality predictor assigns a quality score to each segmentation mask, without using ground truth for the image(s). A set (one or more, but less than all) of the highest quality scores is identified for each image. In some examples, an image processing task is performed using the segmentation masks having an assigned quality score that is within the set of highest quality scores. In some examples, only the segmentation mask having the highest quality score for an image is used in the image processing task. In some examples, a quality threshold is provided, and the segmentation masks meeting the quality threshold are used in the image processing task.

Systems and methods described herein utilize synthetic pixel generation using a custom neural network to generate synthetic versions of objects hidden by occlusions for effective detection and tracking. A computing device stores an object detector model and a synthetic image generator model; receives a video feed; detects objects of interest in a current frame of the video feed; identifies an occluded object in the current frame; retrieves a previous frame from the video feed; generates synthetic data based on the previous frame for the occluded object; and forwards a modified version of the current frame to an object tracking system, wherein the modified version of the current frame includes the synthetic data.

A method and server for training an Object Detector (OD) to detect objects in 3D point clouds are provided. The method comprises: during a first stage of a training pipeline: training the OD using a source domain dataset to detect the objects in a source domain, thereby generating a first trained OD; during a second stage of the training pipeline: training the first trained OD using a target domain dataset to detect the objects in a target domain, thereby generating a second trained OD; and during a third stage of the training pipeline: generating, based on the source domain dataset and the target domain dataset, a cross-domain dataset; and training the second trained OD using the cross-domain dataset to detect objects in both the source domain and the target domain, thereby generating a cross-domain OD.

The invention discloses a data multi-scale fusion method for digital twinning of power transmission and transformation equipment, which relates to the technical field of power equipment and comprises the following steps: acquiring data of the power transmission and transformation equipment by using a laser radar and a camera and calibrating the data to obtain the calibrated data of the power transmission and transformation equipment; performing target detection on the calibrated power transmission and transformation equipment data by using a target detector to obtain a detection result; performing data fusion based on the detection result to obtain an observation set; and inputting the observation set into a BiLSTM network to perform state classification of the power transmission and transformation equipment, and optimizing the BiLSTM network to obtain a state evaluation result of the power transmission and transformation equipment. According to the method, a multi-angle joint calibration technology is adopted, a YOLOv8 target detector is respectively applied to carry out image classification on laser and visual data, a bidirectional long-short term memory BiLSTM model is adopted to carry out quality evaluation on a detected target, and an Adam optimizer is used to carry out hyper-parameter optimization, so that the accuracy of quality detection is improved.

The present invention discloses a small sample target detection method based on adaptive feature generation. According to the number of samples in the data category, the training samples are divided into base class training samples and new class training samples. Based on this, the Faster-RCNN detector is trained and fine-tuned to detect small sample targets. The present invention divides the training set into base class training samples and new class training samples according to the number of data categories in the training samples, effectively improving the detection accuracy of the small sample target detector based on the two-stage strategy, especially after using an adaptive RoI feature generator, the detection accuracy can be further improved. The synthetic features generated by the present invention can align with the real RoI features, effectively overcome the distribution deviation of the synthetic features, achieve accuracy surpassing on all average indicators, and can accurately locate the target object and classify the target object.

The invention discloses a pedestrian re-identification method based on image feature fusion and coding, and the method comprises the steps: constructing a target detector based on a cavity convolution feature pyramid network according to a constructed mixed receptive field module and a low-layer embedded feature pyramid module; classifying and positioning the to-be-detected target images with different scales to obtain divided pedestrian images; a dense feature pyramid network DFPN constructed by a feature extraction module RVNet, a feature enhancement module I DPFM, a feature enhancement module II IRFB and a feature aggregation module Heads is adopted to extract features in a pedestrian image and enhance the features; a pedestrian feature learning method module based on multi-view comparison predictive coding is constructed and embedded into a dense feature pyramid network (DFPN), and multi-view feature reconstruction is performed on input image features by using kernel density estimation and is combined with comparison learning, so that the model learns pedestrian feature expressions of different views. The inter-class difference is increased, and the intra-class similarity is kept; and pedestrian re-identification in the image is realized.

An object detection system that can detect drone objects with high accuracy and low computational complexity, includes a camera for capturing an image, processing circuitry, and a display device. The processing circuitry is configured to input the image. The processing circuitry includes a machine learning network, having a feature extraction backbone with addition-based filters that use addition as a similarity measure to extract features of the image, a path to add low-level features to high-level features, and a single shot detector (SSD) network that outputs an image with possible classes of an object in the image based on the extracted features. The display device displays the image with a label for the detected object based on a selected class. The SSD network backbone can be configured with a SWIN transformer to extract features. The SWIN transformer includes a shifted window self-attention and allows training the SSD model with dynamic image sizes.

An auto-exposure control is proposed for high dynamic range images, along with a neural network for exposure selection that is trained jointly, end-to-end with an object detector and an image signal processing (ISP) pipeline. Corresponding method and system for high dynamic range object detection are also provided.

An auto-exposure control is proposed for high dynamic range images, along with a neural network for exposure selection that is trained jointly, end-to-end with an object detector and an image signal processing (ISP) pipeline. Corresponding method and system for high dynamic range object detection are also provided.

A method for performing machine vision recognition activities for a dynamic object is disclosed, which includes the steps of (i) creating a digital image series that includes a seed digital image set including digital images and at least one subsequent digital image set including digital images; (ii) creating an object detector series, said creating step (ii) comprising a) creating a seed object detector from the seed digital image set, the seed object detector comprising an architecture and a set of weights directed to recognition of target objects; and b) creating at least one deep learning object detector from the at least one subsequent digital image set and derived from said seed object detector, the deep learning object detector including a deep learning architecture and a set of weights trained for recognition of the evolution of the target objects over time.

An auto-exposure control is proposed for high dynamic range images, along with a neural network for exposure selection that is trained jointly, end-to-end with an object detector and an image signal processing (ISP) pipeline. Corresponding method and system for high dynamic range object detection are also provided.