4398 results about "Layers" patented technology

A system and method for adaptive real-time multi-modal compression with dynamic resource allocation provides intelligent compression optimization based on continuously monitored device conditions. The system monitors battery level, CPU utilization, and memory availability while classifying incoming multi-modal data streams comprising image, audio, text, and sensor data to determine processing priorities. Multi-objective optimization balances compression efficiency, reconstruction quality, and energy consumption using evolutionary algorithms that generate optimal parameters for an adaptive variational autoencoder. The autoencoder features dynamically selectable processing complexity, adjustable latent space dimensionality, and modality-specific processing layers. The system automatically switches between operational modes including emergency mode triggered by resource constraints, which applies maximum compression settings and intelligent data triage. Continuous learning adapts compression parameters based on observed performance outcomes, improving future optimization decisions. The system enables homomorphic operations on compressed data and provides enhanced compression performance under varying resource constraints across diverse edge computing applications.

The invention relates to the technical field of defect detection, in particular to a special equipment defect automatic identification method and system, and the method comprises the following steps: marking a continuous path based on an image gray scale gradient and adjacent differences, identifying the pixel connection intensity to obtain a contour image layer, extracting continuous pixels, analyzing gray scale and gradient features, and expanding a texture direction to generate a defect image block. And marking and connecting the small regions to obtain a closed layer, analyzing texture and edge differences, matching classification tags, and clustering and coding to generate a defect identification tag set. According to the method, boundary structure recognition is enhanced through gray gradient and direction continuity analysis, a texture aggregation area is expanded in combination with gray consistency and edge direction stability, brightness abrupt change points are removed, contour extraction precision is improved, micro crack continuity is recovered, and pseudo defects are eliminated; and multi-dimensional image attributes are fused to identify key region feature differences, so that the classification precision and the spatial mapping consistency are improved, and efficient and accurate defect identification and stable classification are realized.

Image series transformation for optimal compressibility is performed with neural upsampling and error resilience. A novel correlation network composed of convolutional layers for feature extraction that extract multi-dimensional features from the image and a channel-wise transformer with attention to capture complex inter-channel dependencies. An angle optimizer enhances compressibility of an image and an error resilience subsystem improves robustness against transmission errors and data loss. The error resilience subsystem applies forward error correction coding, data partitioning based on importance, and embeds error concealment hints. This hybrid approach addresses both local and global features, mitigates compression artifacts, improves image quality, and enhances data integrity during transmission. The correlation network incorporates error correction and concealment techniques during decoding. The model's outputs enable effective image reconstruction, achieving advanced compression while preserving information for accurate analysis.

A method based on a multi-task learning model comprises the following steps: acquiring and processing histopathological image data of gastritis through a data preparation and preprocessing step; feature extraction is performed by using a self-supervised learning pre-trained model, and image blocks are coded into high-dimensional feature vectors; and constructing a multi-task learning model, learning feature representation through a full connection layer module and an attention layer module, and outputting a classification result of each task. And carrying out model training and optimization by using an optimizer, adding multitask loss through a loss function, and dynamically adjusting the model performance. The trained model can automatically detect and grade gastritis, atrophy, acute activity, intestinal metaplasia and other pathological indexes, outputs a standardized evaluation result, and assists in pathological diagnosis. According to the method, a multi-task deep learning framework based on self-supervised learning pre-training is constructed, a traditional single-task modeling mode is broken through, deep learning framework design is driven through pathological index association, and accuracy and clinical practicability of non-neoplastic disease assessment of the stomach are remarkably improved.

An image-guided diffusion network has two Convolution Neural Networks (CNNs). A RGB image and an IR image are concatenated with a Gaussian noise image and input to a denoising neural network that merges information from the RGB and IR images as noise is removed over many iterations. Then an enhancement neural network up-samples for Super Resolution (SR) and convolutes to generate a condition vector that controls Global Feature Modulation (GFM) at three convolution layers to generate a SRGFM enhanced fusion image. Timesteps are embedded using adaptive group normalization blocks within Adaptive Bottleneck Residual (ABR) blocks in the denoising network, which is a UNet having many levels of ABRs, and in the enhancement network before feature modulation. Global image features are detected by triple convoluting the image input to the enhancement network to generate the condition vector that controls feature modulation blocks at three layers of convolution.

The present invention relates to the technical field of deep learning and computer vision. Disclosed is an unmanned aerial vehicle image-based small object detection method for target areas. The present invention crops images of obvious small objects in certain target areas, and annotates the small objects of different categories to form a raw training and testing dataset, so as to ensure the accuracy of data required in the early stage of the algorithm and further ensure the scientificity of the algorithm; uses the computing capability of an improved YOLOv7 detection model to collect image features of different degrees in the dataset, the improved YOLOv7 detection model using YOLOv7 as a basic model and adding to a neck network an MS-CET module, which is constituted by an improved self-attention mechanism and convolution module SPPCSP, and a BHC-FB module, which is constituted by bidirectional mixed convolution modules NConv and RPConv connected in parallel; and finally fuses different feature layers as a final judgment basis of an unmanned aerial vehicle for small object detection in the target areas, to further check the accuracy of the algorithm and criteria for dataset selection, thereby improving recognition accuracy.

The invention discloses an intelligent planning method for space monitoring of an unmanned aerial vehicle. The method comprises the steps that intelligent path allocation is realized by constructing a task demand priority matrix; the method comprises the following steps: firstly, collecting geographical, climate and environmental parameters of a monitoring area, quantifying regional complexity and color features of a monitoring target by combining high-resolution image data with a neural network model, and generating a priority matrix according to task importance, change frequency and risk level; a Dijkstra algorithm is adopted to plan an initial flight path giving consideration to priority and flight limitation, a path complexity index is calculated, and the index comprehensively considers a target priority weight, a task detouring coefficient and a path relaxation degree; and finally, dividing a monitoring area into height layers according to a path complexity index threshold value, performing height layer adjustment on the initial path, and generating a dynamic flight path containing height layer switching, thereby realizing efficient resource allocation and accurate risk prevention and control in a complex monitoring scene.

The invention discloses a coal rock fracture intelligent extraction method based on improved U-Net. The method comprises the following steps: S1, constructing a coal rock fracture CT image data set; s2, constructing an improved U-Net segmentation model, specifically comprising the following steps: S2.1, taking VGG16 as a backbone network, and introducing a depth separable convolution module; s2.2, a PPA attention module is added after each layer of depth separable convolution of the decoder, the PPA attention module is introduced after each up-sampling stage of the decoder, and the output of the PPA attention module is subjected to batch normalization and Dropout layer processing; s2.3, defining a composite loss function; s3, training and optimizing a segmentation model, wherein the specific steps comprise: S3.1, setting hyper-parameters; and S3.2, training the model by using the training set, adjusting hyper-parameters by using the verification set, and evaluating the performance by using the test set, wherein the evaluation indexes comprise MIoU, MAcc and FWIoU. According to the method, the problems of difficult identification of small fractures, large model calculation amount, poor multi-scale information fusion and class imbalance in the coal rock fracture image can be solved, and the robustness, segmentation precision and practicability of the model are improved.

The invention discloses an infrared small target detection method fusing local prior and a multi-scale global background, and the method comprises the steps: firstly obtaining image data containing an infrared image and a mask label corresponding to the infrared image, and carrying out the preprocessing; secondly, a target detection model of an encoder-decoder architecture is constructed, an encoder comprises a local detail prior mining branch and a multi-scale global background perception branch which are parallel, step-by-step feature extraction is performed on the preprocessed image data, and a decoder comprises a progressive feature fusion decoding branch; and inputting the features of each level of the encoder double branches into decoder branches for decoding step by step to obtain a detection result. And finally, a weighted depth supervision mechanism is introduced in training, auxiliary prediction output is set in a plurality of decoding layers, and weighting loss is calculated. According to the method, the problems of insufficient local detail modeling, insufficient multi-scale global background perception of Mamba, difficulty in global and local feature fusion and the like in the existing method are solved, and the detection precision of the infrared small target is improved.

The present invention sets forth techniques for performing style transfer from multiple supplied style images to a supplied content image to generate novel images that include style elements from the multiple supplied style images and content elements from the supplied content image. The techniques include guiding one or more self-attention and cross-attention layers included in a machine learning model based on the multiple supplied style images, such that content elements and style elements included in the style images are not entangled when generating the novel images. The techniques also distill a small subset of representative attention map values from multiple style images, improving performance while reducing computational costs compared to processing all attention map values from the multiple style images.

The invention belongs to the technical field of industrial detection, and discloses a defect detection method and system based on honeycomb catalyst stacking. Omnibearing image data of honeycomb catalyst stacking are obtained through a multi-angle polarization imaging technology, pixel-level polarization degree parameters are calculated to construct a global polarization feature map, and accurate distinguishing between an intrinsic porous structure and suspected defects is achieved. A blind area identification and virtual view angle reconstruction mechanism is introduced, so that the problem of a stacked edge detection blind area is solved; and a layered reflectivity compensation function is adopted, so that the optical interference of an interlayer overlapping region is eliminated. Texture features are extracted through multi-scale morphological filtering, multi-dimensional feature fusion is carried out in combination with polarization features, edge continuity indexes and correction reflection intensity, and a high-precision defect discrimination model is established. And for a low-confidence region, dynamically adjusting detection parameters and performing iterative optimization to form an adaptive detection closed loop. According to the invention, the detection precision and reliability are improved, and the defect position, type and severity can be accurately output.

The invention relates to the technical field of intelligent detection, and discloses an intelligent detection method, device and equipment for a plastic container, and the method comprises the following steps: carrying out multi-angle image acquisition on preforming, blank forming, multi-layer co-extrusion and final blowing forming stages of a plastic container blow molding process to obtain an original image data set; performing targeted preprocessing on the images in each forming stage to obtain an enhanced image data set, and performing feature calculation to obtain multilayer forming feature description data; establishing a feature mapping relationship among different forming stages to obtain defect position and type data; according to the blow molding equipment control system and method, full-process monitoring of the whole blow molding process is achieved, closed-loop control over defect detection and production parameter adjustment is achieved, and the production efficiency is improved. The technical problem that interface defects of all layers in the multi-layer forming process cannot be monitored in real time through a traditional detection method is solved, and the qualification rate of plastic containers is increased.

The invention provides a three-dimensional human body posture estimation method and system based on multi-view visual information fusion and a storage medium, and the method comprises the steps: 1, designing the front half part of a model into Ender Layers with the same layer number as a Transform decoder at a multi-view feature fusion layer, carrying out the data enhancement of an input multi-view original image, and carrying out the reconstruction of the Ender Layers in the multi-view feature fusion layer; inputting the CNN Backbone with the shared weight to extract an initial feature map; 2, introducing a micro-reprojection optimization mechanism, deeply fusing the multi-view geometric consistency constraint into a model training process, and guiding the model to predict a three-dimensional attitude end to end; and step 3, constructing a dynamic projection compensation module. The method has the beneficial effects that the method is particularly suitable for capturing human body posture information in a multi-person interaction scene, the shielding problem and depth estimation ambiguity in a single view angle can be effectively overcome, and the robustness, precision and efficiency of three-dimensional human body posture estimation are remarkably improved.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for fine-tuning diffusion-based generative neural networks in compact parameter spaces for text-to-image generation. In one aspect, a method performed by one or more computers for fine-tuning a diffusion-based generative neural network to obtain a fine-tuned version of the diffusion-based generative neural network is described. The method includes: for each of a number of neural network layers of the diffusion-based generative neural network: obtaining an initial weight matrix including a number of pre-trained weights parametrizing the neural network layer: performing a singular value decomposition on the initial weight matrix; and re-parametrizing the neural network layer with new weights that depend on spectral sifts; and training the spectral shifts of each of the number of neural network layers of the diffusion-based generative neural network to obtain the fine-tuned version of the diffusion-based generative neural network.

The invention discloses a mobile terminal streetscape image real-time segmentation method based on a lightweight neural network, and relates to the technical field of image segmentation. The method comprises the following steps: firstly, carrying out 320 * 320 adjustment, Z-score standardization, adaptive histogram equalization and 3 * 3 Gaussian filtering preprocessing on an input streetscape image; then, an improved MobileNetV3 backbone network is used, and a five-scale feature map is output in combination with DropBlock regularization through eight feature extraction stages including depth separable convolution and an SE attention module; multi-scale features are fused through a U-shaped structure, and a fusion feature map is generated through up-sampling, element-by-element addition of dimension reduction low-layer features and an attention gating module; and during reasoning, outputting a segmentation mask by using a convolutional layer, Softmax and a conditional random field, and finally performing knowledge distillation, weight pruning, 8-bit quantization and TensorRT optimization. According to the invention, high-precision real-time street view segmentation is realized, the robustness is high, and the method is suitable for different devices and scenes.

The invention discloses a self-isolation type image sensing structure, a sensor and a preparation method, and belongs to the field of semiconductors, the self-isolation type image sensing structure is characterized in that a shallow trench isolation structure is exposed after the back surface of a substrate is thinned, then a first semiconductor layer is extended on the back surface of the substrate, and a first isolation structure is prepared in the first semiconductor layer; sequentially depositing a second semiconductor material layer and a metal grating material layer on the first semiconductor layer; sequentially etching the metal grating material layer and the second semiconductor material layer to form a plurality of metal grating structures distributed at intervals and corresponding self-isolation structures; extending a plurality of photosensitive layers in the photosensitive region groove between the adjacent self-isolation structures to form a photosensitive region; and then activating, preparing a filter layer and the like are carried out to obtain a complete back-illuminated image sensor. The self-isolation photodiode is obtained by changing the structural distribution of the photodiode, unexpectedly, the problem that the substrate is damaged by doping is solved, meanwhile, the crosstalk effect is greatly reduced, and the performance of the image sensor is improved.

The invention relates to an underwater image processing technology in the field of underwater autonomous perception, in particular to an underwater low-quality image enhancement method based on a Laplacian pyramid and contrast learning. Comprising the following steps: S1, image input and multi-scale decomposition: decomposing an input image into a low-frequency residual layer and a plurality of high-frequency detail layers through Laplacian pyramid decomposition; s2, inputting the low-frequency residual error layer into a global illumination and color correction sub-module to obtain an enhanced low-frequency residual error layer; s3, inputting the enhanced low-frequency residual error layer and the multi-scale high-frequency detail layer into a frequency domain enhancement feature module to obtain multi-scale enhanced high-frequency detail layer output; s4, performing progressive reconstruction on the enhanced low-frequency layer output and the high-frequency enhancement layer output of each scale according to the inverse process of the Laplacian pyramid; and S5, outputting a result and applying. The method can be used for sharpening optical image data in the operation task of the autonomous underwater vehicle, and the image quality is improved.

The invention provides a visual navigation method and system based on an improved optical flow method, and relates to the technical field of computer vision and inertial navigation. A collaborative optimization framework of an inertial navigation system and visual feature navigation is constructed, carrier motion state parameters are obtained through dead reckoning of an inertial measurement unit, variance parameters of pose variation are calculated, and the layering depth of a pyramid LK optical flow method is dynamically adjusted based on the variance parameters. When the inertial navigation output variance exceeds a preset threshold value, the number of layers of an image pyramid is increased to cope with large-range motion, and otherwise, calculation levels are reduced to improve real-time performance. A dual-period optical flow / feature matching tracking mechanism is designed, improved sub-pixel-level optical flow tracking is adopted in a short period, the calculation complexity is reduced while the positioning precision is guaranteed, feature matching is switched to in a long period, accumulative errors are eliminated, and a period parameter N is determined by inertial navigation precision. According to the method, the calculation complexity can be reduced, and a better effect can be achieved on a low-calculation-performance platform.

An illustrative storage management system is configured to provide persistent data storage to workloads managed by a container orchestration platform deployed in a cluster of nodes. The storage management system is configured to receive a request comprising identifiers indicative of a plurality of layers associated with a virtual machine to be run in the cluster, generate, based on the request, a virtual machine disk image comprising the plurality of layers, the virtual machine disk image configured to be used to run the virtual machine in the cluster, and provide the virtual machine disk image in response to the request. In some implementations, the virtual machine disk image is used by a virtual machine handler associated with the container orchestration platform to run the virtual machine in a virtual machine type pod that is managed by the container orchestration platform.

The invention discloses a medical image classification method and system based on multi-scale spatial state modeling, and the method comprises the steps: firstly dividing an input medical image into a plurality of non-overlapping image blocks, and mapping the non-overlapping image blocks to a feature space through a learnable linear projection layer to obtain an initial feature map; then, multiple layers of stacked MS-SMamba blocks are used for carrying out layer-by-layer feature extraction, each MS-SMamba block comprises a main branch, an auxiliary branch, a dynamic gating fusion network, a residual error connection unit and a feedforward network, and long-range dependency relation capture and multi-scale feature fusion are achieved; and finally, processing the last-layer output feature map through a global feature aggregation and classification module, generating a global feature vector, and outputting a classification result. According to the method, the capturing capability of complex pathological features in the medical image is improved, the calculation efficiency and clinical applicability are improved, and the method is suitable for scenes such as disease screening and auxiliary decision making in medical image diagnosis.

The invention discloses a fan metal surface defect detection method and device based on lightweight YOLO11 and a medium, and relates to the technical field of computer vision and industrial detection. The method comprises the following steps: acquiring fan metal surface defect image data, and preprocessing to obtain a training data set; yOLO11n is used as a basic model, and a lightweight StarNet adopting a four-level layered architecture and a star operation feature fusion mechanism is used as a model backbone network; combining a bottleneck module with a multi-scale convolution block, and constructing a neck network by applying a global heterogeneous kernel selection mechanism and an efficient up-sampling module; a heavy parameterized detail enhanced convolution and group normalization GN strategy is used to construct a detail enhanced lightweight shared convolution detection head; and sequentially connecting the backbone network, the neck network and the output layer of the detection head to form the lightweight metal surface defect detection model. On the premise that the detection efficiency is guaranteed, high-precision identification of metal weld defects can be achieved.

The invention provides a reinforcing steel bar binding point detection method based on improved YOLOv8, and the method comprises the following steps: obtaining a construction site image, carrying out the preprocessing, inputting a to-be-detected image into an improved target detection model, sequentially carrying out the local space attention calculation, depth separable convolution and channel attention weighting operation, and carrying out the detection of the to-be-detected image. Detail feature expression of reinforcing steel bar intersection points is enhanced; performing dynamic weight fusion on the top-down first feature path and the bottom-up second feature path, and embedding a lightweight residual unit in a fusion node; a classification task and a regression task are respectively processed by adopting a classification branch and a regression branch, the classification branch outputs category confidence, and the regression branch outputs coordinate positioning through depth separable convolution and a probability distribution prediction layer; and decoding output data of the target detection model, and generating a detection result containing the position and the category of the binding point. According to the method, the requirement of real-time detection can be met while the accuracy of steel bar binding point detection is improved.

The invention provides a depth estimation enhanced semantic segmentation method and device, equipment and a medium. The method comprises the following steps: extracting a corresponding depth image from an acquired RGB image by using a depth estimation large model; constructing a scene perception model, wherein the scene perception model comprises a coding layer and a decoding layer; the coding layer comprises an RGB image coding branch and a depth image coding branch, and is used for correcting and fusing the characteristics of the RGB mode and the depth mode output by the RGB feature extraction layer and the depth feature extraction layer based on data fusion modules arranged between the coding layers respectively to obtain the first fusion characteristics of each layer; a first fusion feature obtained after fusion of the data fusion modules except the last layer of data fusion module is input to a multi-scale fusion decoding module, step-by-step recovery of a feature map is achieved through a multi-scale feature fusion module and a frequency perception feature fusion device of the multi-scale fusion decoding module, and finally a semantic segmentation result is obtained. According to the method, scene perception is accurate, and meanwhile, only the minimum cost is needed.

The invention discloses a medical image segmentation method and system based on a hierarchical pre-training model, and the method comprises the steps: carrying out the size adjustment of an input image, and carrying out the data enhancement operation; extracting multi-scale features of the medical image by using an encoder based on a hierarchical pre-training visual model, and inserting an adapter in front of each encoding block; a decoder based on a dense connection structure is adopted to gradually recover the spatial resolution, a feature enhancement module is applied behind each decoder block for feature enhancement, and dense connection is achieved through layer-by-layer up-sampling and cross-layer connection; multi-scale supervision and prediction are realized through three output heads, a main output head generates a main segmentation result, and two auxiliary output heads respectively generate segmentation results from an intermediate decoding layer to provide a multi-scale supervision signal. Therefore, the problems of precision, accuracy and efficiency of an existing medical image segmentation technology in tiny focus recognition are solved, and the recognition capability of tiny boundaries and complex geometrical shapes is remarkably improved.

An image processing framework for multi-class classifying a subject into cognitive normal, mild cognitive impairment and Alzheimer's disease (AD) conditions is developed. In one realization of the framework, an AD_Net model, which is an attention-enhanced convolution neural network (CNN) formed by embedding a Convolutional Block Attention Module (CBAM) into a CNN having a Visual Geometry Group 19 (VGG19) architecture, processes an image volume of the subject's brain to generate a plurality of AD_Net feature maps and a first plurality of scores that predict respective likelihoods of the three conditions. To enhance the prediction accuracy, a multilayer perception model formed with a plurality of fully connected layers processes the plurality of AD_Net feature maps and a plurality of influencing factors of AD, such as age, gender, geriatric depression scale score, Mini-Mental State Examination score and clinical dementia rating score, to generate a second plurality of scores that predict the respective likelihoods.

The invention discloses a target detection method based on a YOLO model, electronic equipment and a storage medium, and relates to the technical field of target detection. Comprising the following steps: inputting an aerial image of an unmanned aerial vehicle into a trained target YOLO model; the target YOLO model comprises a backbone network, a neck network and a head network, and a feature extraction module in the backbone network performs multi-scale feature extraction by adopting a double-branch architecture attention mechanism; performing multi-scale feature extraction on the aerial image by adopting a dual-branch architecture attention mechanism through a feature extraction module in the backbone network, and constructing to obtain a plurality of layers of first comprehensive image features of the aerial image; performing feature fusion on the first comprehensive image features of different levels through a neck network to obtain second comprehensive image features of multiple levels; and inputting the multiple levels of second comprehensive image features into a head network to obtain a target detection result of the aerial image. According to the invention, the accuracy of small target detection can be improved.

The invention provides a lightweight visible light ship target detection method based on edge feature guidance, and relates to the technical field of ship detection image data processing, and the method comprises the steps: collecting remote sensing satellite images, and carrying out the random distribution of the images after screening and marking, and obtaining a training set and a verification set; the backbone network module comprises a plurality of Conv modules and C3k2 modules which are mutually stacked; the neck module comprises a detail-enhanced convolution module and a hierarchical pyramid module based on dynamic feature aggregation; in the head module, after the features of all detection layers are subjected to independent convolution processing, feature transformation is carried out through a multi-branch detail enhancement convolution module; performing data enhancement on the training set; and obtaining a trained ship target detection model through a back propagation algorithm and a gradient descent optimization method. According to the invention, the lightweight and precision improvement of the detection head are realized, the robustness of the model to the illumination change is enhanced, and the global semantic information and the local detail features are fused to balance the detection of the small target and the large target.

The invention discloses a method for detecting the quality of a functional layer of a building outer wall by an unmanned aerial vehicle, and particularly relates to the technical field of building outer wall detection.The method comprises the steps that firstly, an infrared image of a building outer wall facing object is collected through the unmanned aerial vehicle, pixel points and remaining pixel points of a serious hollowing area are determined, and the abnormal degree of the remaining pixel points is calculated; clustering is carried out by adopting a defect probability-based region growing method; based on seed point screening of morphological preprocessing, structural elements of different sizes are comprehensively determined to be used through weighted summation calculation according to the jitter frequency of the unmanned aerial vehicle and the image resolution, and noise is filtered step by step. Generating a multi-scale pyramid for the original image, and detecting candidate seed points at each level; sorting and preferentially selecting the candidate seed points based on the morphological closure degree and the shape regularity of each candidate seed point; pixel points in the neighborhood of the growth seed points are merged to obtain all hollow defect connected domains; and judging the quality condition of the building outer wall functional layer according to the area of the hollowing defect connected domain.

The invention provides a small target detection method and system based on an improved YOLOv8 structure, and the method comprises the following steps: carrying out the preprocessing of a to-be-detected image, so as to obtain an input standardized image; inputting the standardized image into a YOLOv8 backbone network, and extracting feature maps of different scales; the feature map is input to a YOLOv8 neck network for feature fusion, the neck network introduces an ASF mechanism, a scale sequence feature fusion module and a three-feature encoder module are integrated, cross-scale enhancement and integration are performed on features of different resolutions, and a multi-scale fusion feature map is generated; constructing P2, P3, P4 and P5 detection layers based on the fused feature map; the P2-P5 detection layer predicts candidate target frames and category confidence of the candidate target frames respectively; soft non-maximum suppression is adopted to process the candidate target frame, a linear or Gaussian attenuation function is utilized to adjust confidence, a redundant frame is suppressed, and a final detection result is output. According to the method, the recognition capability of small targets in dense and complex scenes is remarkably improved.

Innovations in systems, methods, and software for features of a neural image or video codec are described herein. For example, a neural video encoder can receive a current video frame, encode the current video frame to produce encoded data, and output the encoded data as part of a bitstream. As part of the encoding, the encoder can determine a current latent representation for the current video frame, and encode the current latent representation using an entropy model network that includes one or more convolutional layers. As part of the encoding the current latent representation, the encoder can estimate statistical characteristics of a quantized version of the current latent representation based at least in part on a previous latent representation for a previous video frame, and entropy code the quantized version of the current latent representation based at least in part on the estimated statistical characteristics.