168 results about "Prototype learning" patented technology

The invention relates to the technical field of medical image segmentation, in particular to a few-sample medical image segmentation method based on edge perception multi-prototype learning, and the method comprises the steps: inputting support and query images into a feature encoder, and extracting support and query feature maps of different sizes; inputting into a local attention fusion prototype generator to generate a support foreground prototype; processing the support mask through dynamic corrosion operation to generate an inner boundary prototype; generating a multi-foreground local prototype through a multi-layer perceptron; local and global information is optimized through multi-scale feature extraction, and a multi-scale prototype is obtained; fusing to obtain a multi-prototype foreground prototype; dynamic calculation weighting is carried out on the multi-prototype foreground prototype by using a double-stage prototype optimization network, and automatic calibration is carried out; then prediction is carried out through a prototype prediction module, and finally collaborative optimization is carried out through a loss calculation module; the method can effectively solve the problem of edge detail loss involved in the background technology.

The invention relates to the technical field of aircraft intelligent detection and maintenance systems, and discloses an aircraft defect intelligent evaluation system and method based on multi-modal fusion, and the system comprises a multi-modal data collection module, a tensor construction and decomposition module, a meta-prototype relation network module, a multi-target game optimization module, and a closed-loop feedback module. The method comprises the steps of constructing a five-order feature tensor through multi-modal data synchronous acquisition and space-time alignment, extracting low-rank features through hypergraph block item decomposition, dynamically generating a defect prototype set in combination with meta-learning, generating a maintenance decision by adopting a Nash equilibrium strategy and fusing multi-constraint conditions, and optimizing system parameters through closed-loop feedback. The whole-process intelligentization of aircraft defect detection and maintenance is realized; according to the method, high-precision defect detection is realized through multi-modal data fusion and hypergraph modeling, an intelligent decision is generated in combination with dynamic prototype learning and multi-target game optimization, and continuous self-optimization is performed by means of a closed-loop feedback mechanism, so that the operation and maintenance efficiency and safety of the aircraft are improved automatically in the whole process.

The invention provides an unsupervised continuous emotion electroencephalogram analysis method and device based on deep reinforcement learning, and the method comprises the steps: in a prototype learning stage, obtaining sample features of time sequence signals corresponding to sample objects, carrying out the clustering processing of the sample features based on a K-Means algorithm, obtaining the global distribution information of emotion information in the time sequence signals, and carrying out the clustering processing of the global distribution information; in the reinforcement learning stage, a distribution-prototype reward function is determined according to the global distribution information, an Actor-Critic framework is adopted, a near-end strategy optimization algorithm PPO is used, an Actor executes a current action for a given time sequence signal based on a current strategy, Critic determines a current score of the current action according to a distribution-prototype reward fed back by the environment, and the current score of the current action is determined according to the distribution-prototype reward fed back by the environment. And the Actor performs strategy learning based on the current score until a preset iteration termination condition is reached, and a model used for determining key signals in the target time sequence signals according to the target time sequence signals is obtained. Therefore, the accuracy and reliability of the finally obtained model are improved.

The invention provides a model performance optimization method based on heterogeneous federation, which comprises the following steps: a server sends a global model, a global prototype and global knowledge to selected clients, and each client initializes a local model by using the global model; on each client, calculating a loss function by using the global prototype, the global knowledge and a local sample, performing gradient descent, executing a local model obtained after epoch rounds, and calculating a local prototype and local knowledge; after training is completed, the client uploads the local model, the local prototype and the local knowledge to the server. And the server aggregates the local models, the local prototypes and the local knowledge to form a global model, a global prototype and global knowledge, and judges whether a convergence condition or a preset training round number is reached, if yes, training is ended, and if not, the next round of training is executed. The method can effectively solve the problems of inter-class feature stacking, prototype deviation and intra-class feature stacking in a prototype learning technology.

The invention discloses a cross-domain image classification method based on few-sample unsupervised domain adaptation. The method comprises the following steps: acquiring a training feature data set; constructing a feature extractor, and using a pre-training model; building a model based on a sample selection strategy, adversarial domain adaptation and prototype learning; a total loss function is obtained, and iteration is carried out to optimize the model; and performing cross-domain image classification by using the trained domain adaptive classification model. According to the invention, the source domain sample is expanded by using the sample selection strategy without additional unmarked source domain data. Besides, by constructing a source domain category prototype, a target domain sample is directly aligned with the source domain prototype, and refined cross-domain feature alignment is realized. According to the method, local category prototype alignment and global adversarial alignment are combined, the capability of the model in the aspect of feature alignment is enhanced, the excellent classification performance of the model is verified on a public data set, and particularly, remarkable robustness is shown under the condition that source domain labeled samples are extremely limited.

The invention relates to the technical field of machine learning and nuclear power safety assessment, and discloses a nuclear power safety assessment method and system based on machine learning, and the method comprises the steps: collecting and preprocessing the multi-source heterogeneous data of a nuclear power station; constructing a time convolution network model to extract multi-scale time sequence features of the multi-modal data and realize modal alignment; constructing a contrast learning model to optimize multi-modal feature representation; constructing a multi-granularity prototype learning model in the optimized feature representation space; constructing an adaptive attention fusion model to dynamically integrate multi-modal information, and continuously optimizing an evaluation model by a continuous learning technology; through multi-modal fusion and multi-scale time sequence representation learning, comprehensive assessment of the safety state of the nuclear power station is realized, and the comprehensiveness of safety assessment is improved.

The invention belongs to the technical field of artificial intelligence and medical diagnosis, and particularly relates to an interpretable Transform medical diagnosis method based on prototype learning, and the method comprises the following steps: data preprocessing; extracting prototype features; constructing a model; optimizing an attention mechanism; key value pair storage: storing the key value pair as a parameterized prototype; designing a loss function; an Adam algorithm is selected to train the model, and the learning rate and batch size hyper-parameters are adjusted according to the actual situation; and evaluating and optimizing the model. According to the method, a prototype learning mechanism is introduced, the attention mechanism of the Transform model is optimized, and the innovative method overcomes the defect of insufficient model interpretation in the prior art. By combining the prototype features and the attention weight, the model can more accurately position key information, and the accuracy of medical diagnosis is improved.

The invention discloses an improved few-sample medical image target region segmentation method, belongs to the technical field of medical image segmentation, and solves the technical problems of few samples, insufficient intra-class diversity, insufficient inter-class difference learning, poor cross-domain data set segmentation performance and the like in medical image segmentation. The segmentation method comprises the steps of support set processing, prototype production and optimization, cross prototype learning and optimization and query set processing and segmentation, an independent prototype self-attention module and a cross prototype attention module are introduced, fine learning and optimization are carried out on a foreground prototype and a background prototype in a support set image, and a segmentation result is obtained. Therefore, the segmentation capability of the model on the complex medical image under the condition of few samples is enhanced. According to the iterative optimization mechanism, foreground and background prototypes are updated through multiple iterations, the segmentation precision and the robustness of the model are continuously improved, and the method is particularly suitable for processing medical images with high noise and complex backgrounds.

The invention relates to the technical field of artificial intelligence, and discloses an event prediction method and system based on multi-modal fusion, and the method comprises the steps: constructing a knowledge graph encoder, and converting domain expert knowledge into learnable vector representation; constructing a multi-granularity feature extraction network, and extracting features from different microcosmic, mesoscopic and macroscopic scales; realizing a knowledge-guided attention mechanism, and dynamically adjusting feature scale importance; constructing a prototype learning module, and establishing prototype representation of the abnormal category; a knowledge migration mechanism is constructed, and the generalization ability of the model to novel anomalies is enhanced; and multi-granularity abnormal event detection and early warning are realized, and a detection result and interpretable analysis are output. According to the method, through combination of knowledge guidance and multi-scale feature learning, efficient identification and early warning of social abnormal events under the condition of data scarcity are realized, and the method is suitable for the fields of public place safety monitoring, urban traffic safety management, large-scale activity safety guarantee and the like.

The invention discloses an interpretable image recognition method based on visual Transform and prototype learning, and the method comprises the following steps: carrying out the image preprocessing, carrying out the normalization of a to-be-recognized image in training data, dividing the to-be-recognized image into a plurality of non-overlapping image patches, and flattening each image patch into a one-dimensional vector; using different types of visual Transform encoders to extract the long-range dependent feature representation of the image patch; prototype branches are designed, and a double-branch prototype layer and a single-branch prototype layer are formed; according to the type of a visual Transform encoder, through a double-branch prototype layer or a single-branch prototype layer, calculating similarity and converting to obtain an activation graph, and processing the activation graph; constructing a linear classifier to obtain an image recognition result; and explaining the acquisition process of the image recognition result to complete interpretable image recognition based on visual Transform and prototype learning.

The invention relates to a cross-bearing single sample intelligent diagnosis method based on cognitive guidance and Riemannian manifold, and belongs to the technical field of rotating machinery fault diagnosis. Aiming at the problems of insufficient global task distribution learning ability, small sample over-fitting, Euclidean modeling limitation and the like of the existing meta learning method in a cross-domain single-sample scene, a cognitive guidance Riemannian meta learning framework is provided. According to the technical scheme, the method comprises the following steps: 1) constructing cognitive prototype learning global task distribution, and guiding a model to extract high-quality general meta-knowledge from multiple tasks; 2) designing a cognitive adaptive factor to dynamically adjust source domain memory, enhancing target domain adaptation and reducing single sample deviation; and 3) introducing a Riemann metric driving strategy, mapping the data to a Grassmann manifold space, and enhancing the non-linear feature discrimination ability by using geodesic distance. According to the method, the average diagnosis accuracy in a cross-bearing single sample task reaches 93.46% and is improved by 10.04% compared with an existing optimal method, and the accuracy and generalization ability under complex working conditions are improved.

The invention relates to the technical field of pathological image analysis and mining, and particularly discloses an ovarian cancer subtype classification method based on prototype learning and multi-view deep embedding clustering, and the method comprises the following steps: S1, collecting a tissue pathological image of an ovarian cancer patient and a corresponding full-view digital pathological image; and S2, generating a multi-view data set. According to the ovarian cancer subtype classification method based on prototype learning and multi-view deep embedding clustering, the problem that in the prior art, patch-level labels are generally lacked in the field of multi-instance pathological images, so that many natural image processing methods cannot be applied to the field of pathological images is solved. A ResNet backbone network is used for extracting features of pathological images under the maximum magnification, a small number of pathology prototypes are introduced to guide deep embedded clustering through pathology expert priori knowledge, a pathology image spectrogram is introduced to serve as a reference view, and the accuracy and stability of clustering are enhanced.

The invention provides an interpretable fine-grained image recognition method, device and equipment and a medium, and relates to the technical field of fine-grained image recognition, and the method adopts a double-branch structure frame, and combines feature fusion and prototype learning to carry out fine-grained image classification. Wherein the feature extraction layer captures a complex mode and long-distance dependence in an image through a self-attention mechanism, meanwhile, a feature selection fusion module is introduced, information marks of prototype learning are effectively selected, and it is ensured that the learned prototype pays more attention to the prospect. The prototype learning layer designs a double-branch structure, processes and learns a global prototype and a local prototype of an image respectively, improves the accuracy of the network, and explains a model classification result from a global perspective. And the full connection layer realizes a final classification decision, and obtains a final classification probability through weighted sum. The objective of the invention is to solve the problem of black box characteristics of an existing deep learning model in fine-grained image classification and improve the application credibility of the model in the high-risk field.

The invention belongs to the technical field of image segmentation, and relates to a semi-supervised pancreatic image segmentation method based on prototype estimation and prototype consistency. According to the method, based on a semi-supervised segmentation framework of an average teacher, V-Net is adopted as a deep learning segmentation model, and a projection head is inserted in the last but one stage of a decoder as a prototype branch; prototype learning design prototype estimation is introduced, unmarked data prediction is divided into determined areas of fuzzy areas, and different losses are designed for areas with different reliability of the unmarked data. The problems that in an existing semi-supervised learning segmentation method, unmarked data cannot be fully utilized, potential information of marked data is insufficient in utilization and the like are effectively solved, and a higher-performance solution is provided for an abdominal CT image pancreatic organ segmentation task.

The invention provides an image text description generation method, electronic equipment and a readable storage medium. According to the method, the object perception prototype learning module and the global context feature extraction module are introduced, so that fine-grained information and global semantic understanding in the image are effectively balanced. The visual backbone network module can extract multi-scale and multi-level image features and perform fusion, thereby enhancing the expression ability of the image features. The object perception prototype learning module further extracts an object prototype from the fusion features to ensure that the model can accurately capture key objects and attributes thereof in the image, and the global context feature extraction module ensures that the overall context of the image is fully understood. On the basis, the encoding and decoding module combines the global context and the object prototype to generate the text description, so that the semantic splitting phenomenon in the traditional method is avoided, and the detail information in the image is effectively reserved, thereby improving the accuracy and integrity of the image description.

The invention belongs to the computer vision technology, and particularly relates to a small sample multi-modal prototype learning-oriented remote sensing scene classification method, which comprises the following steps of: acquiring RGB (Red, Green and Blue) images with category labels and text prompts of the RGB images as a support set; establishing a text prototype, an RGB prototype and a hyperspectral prototype of each category according to the support set; and extracting to-be-classified query set image features by using a pre-trained CLIP image encoder, calculating cosine similarities between the query set image features and the text prototype, the RGB prototype and the hyperspectral prototype of each category of the support set, taking the cosine similarities as input of a multi-layer perceptron, and obtaining the category of the to-be-classified RGB image through classification of the multi-layer perceptron. High-precision and high-robustness remote sensing scene classification is realized under the small sample condition, only prototype and similarity calculation is needed in the reasoning stage, and deployment and expansion are easy.

The invention relates to a modal heterogeneous federated learning privacy protection method based on a cross-modal prototype, and belongs to the technical field of artificial intelligence and information security. The server firstly initializes a global model for each client, wherein the global model comprises a mapping module of a corresponding mode of the client; and the server sends the initialized model to the corresponding client. The multi-mode client initializes a private prototype learning model. And local model training is carried out by using the global model, the single-mode client and the multi-mode client. And after the local prototype and the local model sent by the client are received, the server executes prototype aggregation and model aggregation respectively. And after receiving the global prototype pair and the global mapping module, the client updates the local mapping module and starts a new round of training until a specific number of training rounds is reached. According to the method and the device, the communication overhead can be reduced, local models are not required to have the same structure, and a single-mode client can acquire information from a missing mode prototype in a targeted manner.

The invention discloses a water-cooled permanent magnet coupler heat dissipation fault diagnosis method based on a neural network, relates to the field of coupler heat dissipation fault diagnosis, and obtains a three-dimensional temperature field with a fault tag and a fault sensitive tensor through multi-physical field coupling and fault tag creation analysis based on an initial three-dimensional temperature distribution field and discrete electromagnetic field data. And performing multi-field fusion and space-time compression on the three-dimensional temperature field with the fault tag and the fault sensitive tensor to obtain a feature vector of concentrated fault key information, and performing physical constraint dimension reduction, fault prototype learning and parameter inversion on the three-dimensional temperature field with the fault tag and the feature vector of the concentrated fault key information to obtain fault parameter estimation. Bidirectional recursive fault analysis is performed based on fault parameter estimation to obtain a propagation state, gradient analysis and adversarial diagnosis are performed on the propagation state to obtain a fault type corresponding to the maximum probability, and misjudgment under complex working conditions can be greatly reduced.

The invention provides a fine-grained three-dimensional model classification method and system based on dynamic prototype learning, and belongs to the technical field of three-dimensional geometric analysis. Comprising the steps of performing multi-view orthogonal projection rendering on a three-dimensional model, performing feature extraction and projection operation on an image sequence of each view after projection rendering based on an encoder, and fusing multi-view features; a shared prototype pool is constructed, a prototype cost matrix is generated, a structured measurement space is constructed by calculating the cosine similarity of the multi-view features and the prototype cost matrix, and dynamic soft allocation is executed; sequentially executing a dynamic updating strategy and joint loss optimization on the shared prototype pool; and measuring the distance between the test sample features and the prototypes in the shared prototype pool to realize the probabilistic decision-making of the fine-grained category. Therefore, a dynamic prototype learning framework with geometric perception capability is constructed, and the precision bottleneck of a fine-grained classification task can be broken through by analyzing interpretable features and prototype mapping relations.

The invention discloses an infrared small target detection method based on semi-supervised multi-view prototype learning, and the method comprises the steps: firstly constructing an infrared motion small target detection network based on a teacher-student network, designing a bidirectional time motion sensor, and fully simulating a motion normal form of a target through forward and backward sensing, and generating a multi-view feature prototype as a soft label to explore a position label, and in combination with target perception motion reconstruction loss, achieving the purpose of detecting a mobile infrared small target on a large-scale unmarked video data set. The method shows higher accuracy and robustness in the aspect of infrared motion small target detection, can effectively detect the small target appearing in the infrared motion small target sequence image collected by the infrared imaging equipment, and has wide application prospects in the important application fields of unmanned aerial vehicle supervision, security and protection and the like.

The invention provides a cancer survival prediction method based on a pathological image, and relates to the technical field of artificial intelligence and medical image analysis. The method comprises the following steps: firstly, dividing an acquired pathological image to obtain a plurality of image blocks; and obtaining the feature vector of each image block and identifying the tissue type to which each image block belongs. Calculating spatial proximity, feature similarity and tissue type compatibility between the image blocks according to the center coordinates of the image blocks, the feature vectors and the tissue type to which each image block belongs; and based on a calculation result, constructing a dynamic heterogeneous graph, and performing feature extraction to obtain a multi-scale feature graph of the dynamic heterogeneous graph. And performing multi-prototype learning on the basis of the belonging organization type of the image block and a cross-category attention mechanism to generate multiple prototypes. And finally, the multi-scale feature map and the multiple prototypes are aggregated to obtain final image representation, and cancer survival prediction is performed based on the final image representation, so that the prediction accuracy of the cancer survival state is improved.

The invention belongs to the technical field of image classification, and relates to a prototype learning-based class increment image classification method, which comprises the following steps of: obtaining image data, and constructing a first data set and a second data set; the first data set and the second data set respectively comprise different sample pairs; constructing an initial model, and training the initial model by adopting the first data set to obtain a teacher model; obtaining sample features of the first data set according to a teacher model; obtaining prototype learning loss according to sample features of the first data set; obtaining the total loss of the model according to the prototype learning loss; based on the total loss of the model, training the teacher model by using the second data set to obtain a student model; and classifying the class increment images by adopting a student model. By adopting the method, the problem of disastrous forgetting can be effectively relieved, and meanwhile, the stability and generalization ability of the model are improved.

The invention discloses a multi-modal Hash retrieval method, system and equipment based on a multi-view center structure and a medium, belongs to the technical field of artificial intelligence and multi-modal Hash retrieval, and aims at achieving the technical purpose of obtaining balance between intra-class compactness and inter-class separability in the multi-modal Hash retrieval process. According to the technical scheme, the method comprises the following steps: constructing a multi-modal data set; constructing a multi-modal Hash retrieval model based on a multi-view center structure; training the model; wherein the construction of the multi-modal Hash retrieval model based on the multi-view center structure specifically comprises the following steps: modal specific prototype learning: respectively extracting refined features from corresponding modals by using an image modal depth multilayer perceptron and a text modal depth multilayer perceptron; calculating the average value of the refined features of the image modality and the text modality to obtain a modality specific prototype, and further obtaining the unique features of the image modality and the unique features of the text modality; carrying out multi-modal integrated prototype learning; and performing multi-view semantic enhanced hash learning.

The invention discloses a remote sensing target detection method for an open set scene, and belongs to the technical field of open set target detection. The open set remote sensing target detection network introduces a cascade target positioning network, a background feature enhancement module and reciprocal point prototype learning on the basis of a Faster-RCNN model; the cascade target positioning network can effectively adapt to multi-scale features, calculates the quality of a candidate box through centrality loss, and replaces a foreground-background classification mode in a traditional region suggestion network, so that the recall rate of objects is increased, and the possibility that unknown objects are misjudged as backgrounds is reduced; the background feature enhancement module fuses various background region information so as to enhance the adaptability to a complex environment and reduce the interference of background noise on target recognition; according to reciprocal point prototype learning, a reciprocal point prototype of a known category and a background is constructed, and an embedded network metric learning strategy is combined, so that background information and the known category jointly participate in candidate box classification, and a known target, an unknown target and the background are effectively distinguished.

The invention belongs to the technical field related to fault signal processing, and discloses a multi-source-domain ball screw fault diagnosis method and system for partial source domain data missing. The method comprises the following steps: carrying out data complementation on vibration signals with partial fault category missing under different working conditions in the operation process of the lead screw, and obtaining complete source domains corresponding to all the working conditions; the source domains and the target domains are input into a domain adaptive network for training, the domain adaptive network comprises a prototype learning module and a classifier, and in the prototype learning module, a minimum maximum mean value difference loss function and a feature norm loss function are adopted to close the feature distribution distance between each source domain and the target domain; the classifier takes the fault type as output; and inputting a to-be-diagnosed vibration signal into the classifier of the domain adaptive network to obtain a fault type corresponding to the to-be-diagnosed vibration signal. Through the method, the problem that an existing diagnosis method cannot diagnose unknown working condition faults under the condition that partial working condition training data are missing is solved.

The invention discloses a fine-grained feature matching vehicle re-identification method for a license plate shielding scene, and the method comprises a vehicle detection stage based on illumination perception and adaptive prediction, and a fine-grained Transform feature matching stage based on implicit semantic prototype perception. Comprising an illumination sensing module and an adaptive prediction module. The illumination sensing module is used for realizing image adaptive enhancement under different illumination conditions; the adaptive prediction module suppresses a redundant prediction frame through a self-attention weight mechanism; in the fine-grained Transform feature matching stage based on implicit semantic prototype perception, an implicit semantic prototype learning module and a fine-grained Transform matching module are included; the implicit semantic prototype learning module unsupervised decouples vehicle structural components through a group of learnable semantic prototypes to generate a soft semantic distribution diagram; and the fine-grained Transform matching module is used for introducing semantic affinity bias in cross attention. According to the method, the accuracy and stability of vehicle detection and identity matching in a complex environment can be improved.

The invention relates to a multi-degradation medical image unified fusion method based on degradation prototype learning, and belongs to the field of medical image fusion. The method comprises the steps that low-dose PET data, CT metal artifact data and MRI data with motion artifacts are generated through the imaging principle; learning a degradation prototype by using a feature selection mechanism; the basic fusion model is decomposed into a plurality of branches through a low-rank decomposition strategy, and processing can be performed through different branches when different degradation data are fused; a prompt module based on a learnable feature prototype is designed, and fusion is promoted by injecting degradation-related invariant features into different LoRa branches; and constructing a fused image through an output layer by integrating the degradation elimination fusion features in different scales. According to the method, the medical images containing degradation can be effectively fused, and the robustness and practicability in reality are improved.

The invention discloses a panoramic activity recognition method based on adaptive two-way propagation prototype learning, and the method achieves the combined recognition of multi-granularity activities in a panoramic scene through an adaptive focusing detector and a two-way propagation prototype learning device. In a self-adaptive focusing detector, individuals are effectively detected through a panoramic adaptive focuser in a coarse and fine combination mode, and changes of individual sizes and shielding in a crowded panoramic scene are dealt with. In a two-way propagation prototype learning device, closed-loop interaction and information consistency among different granularities are promoted through the two-way propagation prototype learning device, and information loss caused by inaccurate positioning is reduced. The method comprises a detection branch and an identification branch, the detection branch outputs an original detection frame and inputs the original detection frame into a panoramic adaptive focuser to obtain a final size adaptive detection result, and the identification branch encodes a video and obtains different levels of feature representations for identification through a bidirectional propagation prototype device. And the overall framework is subjected to joint optimization through detection loss and multi-granularity identification loss.

The invention discloses a weak supervision medical image segmentation method based on MedSAM, and aims to solve the problems that traditional medical image segmentation depends on a large amount of annotation data, and training is time-consuming and labor-consuming, and improve the segmentation performance under the condition that the annotation cost is limited. A deep learning model feature extraction capability is utilized, and a category activation graph is generated through a backbone network to preliminarily position a target area. A CSD bidirectional common sense defense module and an AMS class activation graph superposition module are provided, the former is combined with global information to improve the robustness of the model, and the latter is used for determining the maximum target range to guarantee the segmentation effect. A dynamic threshold segmentation strategy and a MedSAM model are combined to enhance a thermodynamic diagram, semantic features are integrated to avoid dependence on prototype learning, and weak supervision high-precision segmentation is realized. Experiments show that the multi-data-set performance of the method is superior to that of the existing method, the application prospect is wide, and the method can be applied to various medical image tasks and other fields which are high in labeling cost and need efficient segmentation.

The invention relates to a remote sensing image semantic segmentation method and system based on prototype learning, which dynamically optimizes feature representation through prototype learning, captures multi-scale context information in combination with a mask Transform, and significantly improves segmentation precision while keeping model efficiency, especially for a segmentation effect of small targets and complex scenes.