Visual perception model construction and efficient evolution method for UAV aerial images

Abstract

This invention provides a method for constructing and efficiently evolving a visual perception model for drone aerial imagery. The method comprises the following steps: obtaining a dataset for segmenting drivable regions, performing data preprocessing and data partitioning; constructing a lightweight fine-tuning network consisting of a ViT-Adapter and a VPT; building an auxiliary task guidance framework that adds multiple lightweight auxiliary decoders to the original "encoder-decoder" architecture to perform the auxiliary task of "vehicle segmentation"; employing a backpropagation algorithm to iteratively update and optimize network parameters until the model loss converges; and inputting the drone aerial imagery to be segmented into the trained model, with the model output being the segmentation result. This invention overcomes the shortcomings of existing efficient fine-tuning methods, assisting a general image segmentation model in efficiently evolving its line of sight in the task of segmenting drivable regions for drone aerial imagery, thereby improving the performance of the final model in image segmentation perception tasks.

Description

Technical Field

[0001] The present invention belongs to the technical field of basic perception large-scale model construction and evolution, and in particular relates to a visual perception model construction and efficient evolution method for unmanned aerial vehicle aerial images. Background Art

[0002] The task of drivable area segmentation is one of the important tasks in drone aerial photography scenarios. It has important applications in intelligent transportation, military reconnaissance and other fields, and is also a current academic research hotspot. Drivable area segmentation falls under the category of semantic segmentation. In the field of semantic segmentation, "pre-training + full parameter fine-tuning" has become the mainstream learning paradigm. However, the parameter scale of the currently popular basic perception models such as SAM (Segment Anything Model) is usually very large, and the computational complexity is very high when fine-tuning the full amount, and the applicability is poor. To this end, researchers proposed a parameter-efficient fine-tuning method (Parameter-EfficientFine-Tuning, PEFT) to help the model evolve quickly and efficiently.

[0003] The core idea of efficient parameter fine-tuning methods is to fix most of the pre-trained parameters and fine-tune only a small number of the model's original parameters, or fine-tune a small number of additional parameters, thereby minimizing computational and storage requirements. Currently proposed efficient parameter fine-tuning methods can be roughly divided into three categories: selective methods, incremental methods, and parameter-heavy methods. Furthermore, incremental methods can be further divided into two subcategories based on the location of parameter addition: adapter methods and soft-hint methods.

[0004] Selective methods simply update some of the pre-trained model's original parameters. Typical methods include BitFit, DiffPruning, and SPT. BitFit is the most popular of these methods. This method fine-tunes only the bias parameters in the model, making it a simple and effective approach.

[0005] Adapter methods fine-tune a frozen pre-trained model by adding new trainable network layers. A representative example, the Adapter method, adds two adapter network modules to each Transformer layer to fine-tune the pre-trained model. This method has been widely used in natural language processing and computer vision, and a number of variants have emerged, including AdaMix, Scaled Parallel Adapter, and AdapterFormer.

[0006] Soft prompting methods add trainable tokens to the model input. These tokens indirectly adjust the attention mechanism, helping the model focus on features most relevant to downstream tasks, thereby achieving fine-tuning. Typical soft prompting methods in natural language processing include Prompt Tuning, Prefix Tuning, P-Tuning, and P-Tuning v2. Jia et al. applied the soft prompting concept to the field of computer vision and proposed Visual Prompt Tuning (VPT).

[0007] The core idea of reparameterization methods is to first add additional parameters, which can then be incorporated into the model in some way after fine-tuning, effectively optimizing the model directly. The most famous reparameterization method is the LoRA method proposed by Hu et al. This method represents the increment of the weight matrix as the product of two low-rank matrices. Subsequent work has improved LoRA, including KronA and AdaLoRA. The recently proposed SSF also falls into the category of reparameterization methods. This method uses trainable scaling and offset parameters to adjust the weights, which can be incorporated into the original parameters after training.

[0008] Existing efficient fine-tuning methods are designed for versatility, aiming to fine-tune models for a variety of tasks, such as image classification, detection, segmentation, and generation. Few are specifically designed and optimized for a specific fine-tuning task. Existing efficient fine-tuning methods face the following two major challenges when aligning general pre-trained models with the semantic segmentation task of drone aerial photography scenes.

[0009] On the one hand, existing studies have shown that multi-scale feature maps have a positive impact on semantic segmentation tasks. However, general pre-training models based on the Transformer architecture, such as SAM, are mostly "straight-tube" structures that can only generate single-scale feature maps and contain relatively scarce image feature information. Although multi-scale feature maps can be indirectly obtained through upsampling, they are still based on the transformation of features at a single scale, and the image information obtained is still not rich enough. To ensure universality, the currently popular parameter efficient fine-tuning methods only perform fine-tuning within the model by "unfreezing some parameters" or "adding some parameters", making it difficult to change the output scale of the model, limiting the alignment effect.

[0010] On the other hand, in the task of drivable area segmentation in drone aerial imagery, traffic-related scene information, such as road vehicles, can often enhance drivable area segmentation accuracy. However, general pre-trained models are trained on general task datasets and fail to exploit traffic scene-related information. Furthermore, existing efficient parameter fine-tuning methods fail to consider how to help the model understand traffic scenes during the fine-tuning process, resulting in poor segmentation performance on related tasks.

[0011] Therefore, making full use of multi-scale feature maps and introducing traffic scene related information is an effective means to improve the evolution effect of general image segmentation models such as SAM in drivable area segmentation tasks. Summary of the Invention

[0012] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0013] The present invention provides a method for constructing and efficiently evolving a visual perception model for drone aerial images, characterized in that the method comprises:

[0014] S1: Obtain the drivable area segmentation dataset from the drone aerial perspective, complete data preprocessing, and perform data segmentation;

[0015] S2: Constructing a lightweight fine-tuning network: The lightweight fine-tuning network consists of two parts: the Adapter network and the VPT network. The Adapter network includes the SPM module, the Injector module, and the Extractor module. The SPM module is used to generate multi-scale feature maps, and the Injector module and the Extractor module are used to interact with the generated multi-scale feature maps and the features in the SAM image encoder. The VPT network is used to add trainable tokens to the input of each layer of the Transformer model to indirectly adjust the attention mechanism.

[0016] S3: Build an auxiliary task guidance framework: Add multiple lightweight auxiliary decoders to the original "encoder-decoder" architecture to perform the "vehicle segmentation" auxiliary task; during training, multiple decoders share a common encoder and perform decoding separately; finally, the losses of these decoders are weighted summed to obtain the final loss;

[0017] S4: Train the segmentation network: Using the loss function set in S3 and the back-propagation algorithm, iteratively update and optimize the network parameters until the model loss converges;

[0018] S5: Segment the drivable area: pre-process the drone aerial image to be segmented according to the pre-processing method in S1, and then input the pre-processed image into the trained segmentation network. The output result of the segmentation network model is the drivable area segmentation result.

[0019] The technical solution of the present invention provides a method for constructing and efficiently evolving a visual perception model for drone aerial imagery. This method preprocesses and partitions the data, constructs a lightweight fine-tuning network, builds an auxiliary task guidance framework, trains a segmentation network, and finally performs drivable area segmentation. By fully mining multi-scale feature maps and leveraging traffic scene-related information such as vehicle targets, this method enables efficient evolution of a general image segmentation model to the task of drivable area segmentation in drone aerial imagery, improving the model's performance on this task. Compared with the prior art, the technical solution of the present invention can address the technical problems in the prior art, such as the difficulty of directly acquiring multi-scale feature maps for straight-tube general image segmentation models and the difficulty of effectively incorporating traffic scene information into existing fine-tuning methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are included to provide a further understanding of the embodiments of the present invention, constitute a part of the specification, illustrate the embodiments of the present invention, and together with the description, explain the principles of the present invention. Obviously, the drawings described below are only some embodiments of the present invention, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0021] Figure 1 A schematic diagram of a process flow of an efficient evolution method of a general image segmentation model for drivable area segmentation provided according to a specific embodiment of the present invention is shown;

[0022] Figure 2 A schematic diagram of a lightweight fine-tuning network structure for multi-scale feature extraction provided according to a specific embodiment of the present invention is shown;

[0023] Figure 3 A schematic diagram of the auxiliary task guidance framework structure provided according to a specific embodiment of the present invention is shown. DETAILED DESCRIPTION

[0024] It should be noted that, in the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is in no way intended to limit the present invention and its application or use. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0025] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0026] Unless otherwise specifically stated, the relative arrangement of the parts and steps, numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present invention. Meanwhile, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to actual proportional relationships. Technology, methods and equipment known to those of ordinary skill in the relevant art may not be discussed in detail, but in appropriate cases, the technology, methods and equipment should be considered as a part of the specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values.

[0027] like Figure 1 As shown, according to a specific embodiment of the present invention, a method for constructing and efficiently evolving a visual perception model for drone aerial images is provided, which specifically includes:

[0028] S1: Obtain the drivable area segmentation dataset from the drone aerial photography perspective, complete data preprocessing, and perform data segmentation.

[0029] S2: Build a lightweight fine-tuning network. The lightweight fine-tuning network consists of two parts: the Adapter network and the VPT network. The Adapter network includes the SPM module, the Injector module, and the Extractor module. The SPM module is used to generate multi-scale feature maps, while the Injector module and the Extractor module are used to interact with the generated multi-scale feature maps and the features in the SAM image encoder. The VPT network is used to add trainable tokens to the input of each layer of the Transformer model, indirectly adjusting the attention mechanism.

[0030] S3: Build an auxiliary task guidance framework. Multiple lightweight auxiliary decoders are added to the original encoder-decoder architecture to perform the auxiliary task of vehicle segmentation. During training, multiple decoders share a common encoder and perform decoding independently. Finally, the losses of these decoders are weighted summed to obtain the final loss.

[0031] S4: Train the segmentation network. Using the loss function set in S3 and the backpropagation algorithm, iteratively update and optimize the network parameters until the model loss converges.

[0032] S5: Segment the drivable area. The drone aerial image to be segmented is preprocessed according to the preprocessing method in S1, and then the preprocessed image is input into the trained segmentation network. The output of the segmentation network model is the drivable area segmentation result.

[0033] This configuration approach provides a method for constructing and efficiently evolving a visual perception model for drone aerial imagery. This method preprocesses and partitions the data, constructs a lightweight fine-tuning network, builds an auxiliary task guidance framework, trains the segmentation network, and finally performs drivable area segmentation. By fully mining multi-scale feature maps and leveraging traffic scene-related information such as vehicle targets, this method enables efficient evolution of general image segmentation models to the task of drivable area segmentation in drone aerial imagery, improving the model's performance on this task. Compared to the prior art, the technical solution of the present invention addresses the technical issues in the prior art, such as the difficulty of directly acquiring multi-scale feature maps for straight-tube general image segmentation models and the difficulty of effectively incorporating traffic scene information into existing fine-tuning methods.

[0034] Furthermore, in the present invention, S1 is first executed: a drivable area segmentation dataset under the aerial photography perspective of a UAV is obtained, data preprocessing is completed, and data division is performed.

[0035] In this invention, data preprocessing includes label processing and image enhancement. Label processing means retaining only the Road label in UAVid and classifying the rest of the labels as background labels. Image enhancement includes resizing, mosaic enhancement, random cropping and random flipping, etc.

[0036] The data is divided based on the criteria given by the selected dataset.

[0037] Furthermore, in the present invention, S2 is executed to construct a lightweight fine-tuning network. The lightweight fine-tuning network includes two parts: the Adapter network and the VPT network. Among them, the Adapter network includes an SPM (Spatial Prior Module) module, an Injector module, and an Extractor module; the SPM module is used to generate a multi-scale feature map, and the Injector module and the Extractor module are used to interact the generated multi-scale feature map with the features in the SAM image encoder. The VPT network is used to add trainable words at the input of each layer of the Transformer model to indirectly adjust the attention mechanism. The lightweight fine-tuning network structure is as follows: Figure 2 shown.

[0038] In this paper, a lightweight fine-tuning network is built on top of the SAM image encoder. Since the original SAM task header cannot predict semantic labels and cannot be directly used for semantic segmentation tasks, the decoder in this paper uses the UPerNet semantic segmentation task header instead.

[0039] In the present invention, the specific structures and functions of the three components of the Adapter network are as follows:

[0040] The SPM module adopts a convolutional neural network structure, which includes a Stem network and three convolution layers, and will obtain four downsampling scales, namely Then, These three multi-scale feature maps are flattened and concatenated to obtain the feature Input to the following series of Injector modules and Extractor modules. Among them, represents the i-th multi-scale feature containing spatial prior information, Represents a real-valued feature space of a specific dimension, D represents the number of channels of the feature, and HW represents the length-width product of the image.

[0041] The Injector module uses the cross attention mechanism to integrate multi-scale features The spatial prior information contained in is injected into the Transformer Block corresponding to ViT. Specifically, for the i-th Injector module, the input of the i-th Block As Query, multi-scale features As the key (Key) and value (Value), perform attention calculation, and compare the result with the original Add them together and get This enables the injection of spatial prior information. The corresponding formula is shown below.

[0042]

[0043] in, is the input of the i-th Block, is the input of the updated fusion spatial prior information, γ i is a learnable vector used to balance and attention calculation results.

[0044] The Extractor module consists of a cross attention (Attention) and a feedforward neural network (FFN), which uses the previous multi-scale features And the output of the corresponding Transformer Block in ViT Generate new multi-scale features Specifically, for the i-th Extractor module, the multi-scale features As a query, take the output of the i-th Block As the key and value for attention calculation, the result and the original Add together to get Then, Projected through a feedforward neural network to obtain new multi-scale features

[0045]

[0046] in, To fuse multi-scale features and the output of the i-th Transformer Block of ViT The new features obtained, It is the multi-scale feature output by the i-th Extractor.

[0047] The 12 layers in the SAM image encoder are divided into 4 groups, each of which interacts with a pair of Injector-Extractor. After multiple iterations, high-quality multi-scale features (with Three scales). Then, it is combined with the spatial prior module to obtain The scale feature maps are combined to form feature maps with four scales. Finally, the multi-scale features obtained in the Adapter network are fused with the features in the SAM image encoder to obtain a more informative multi-scale feature, which is then passed to the decoder for decoding.

[0048] Furthermore, in the present invention, S3 is executed to construct an auxiliary task guidance framework. Multiple lightweight auxiliary decoders are added to the original "encoder-decoder" architecture to perform the "vehicle segmentation" auxiliary task. During training, multiple decoders share one encoder and perform decoding separately. Finally, the losses of these decoders are weighted and summed to obtain the final loss. The auxiliary task guidance framework is as follows: Figure 3 shown.

[0049] In this paper, the auxiliary task head is built on top of the UPerNet network in the main task head. The auxiliary task decoder directly uses the fused features from the UPerNet network as input, allowing the auxiliary task decoder to use only a simple FCN task head. This eliminates the need for the auxiliary task head to also use UPerNet, reduces the number of trainable parameters, and helps the model evolve more efficiently.

[0050] The labels corresponding to the auxiliary tasks can be obtained directly from the selected dataset (if the dataset has corresponding annotations), or they can be generated using existing tools (if the dataset does not have corresponding annotations). Among them, the "vehicle segmentation" auxiliary task can use the InternImage model to generate auxiliary segmentation maps.

[0051] As a specific embodiment of the present invention, the loss function used for both the main task and the auxiliary task is the cross-entropy loss function. The final loss function is the weighted sum of the loss functions of each task. For example, the loss weight of the main task is 1.0, and the loss weight of the vehicle segmentation auxiliary task is 0.1.

[0052] In response to the technical defects of existing efficient fine-tuning methods, the present invention proposes a method for constructing and efficiently evolving a visual perception model for UAV aerial images. This method can solve the problems that the straight-tube general image segmentation model is difficult to directly obtain multi-scale feature maps, and that the existing fine-tuning methods are difficult to effectively introduce traffic scene information. By fully mining the multi-scale feature maps and utilizing traffic scene-related information such as vehicle targets, the general image segmentation model can be efficiently evolved to the task of segmenting the drivable area of UAV aerial images, thereby improving the performance of the model on this task.

[0053] The beneficial effects of the present invention include:

[0054] (1) A “lightweight fine-tuning network for multi-scale feature extraction” is proposed. This network is built based on ViT-Adapter and VPT, which enables it to extract multi-scale features with rich information while fine-tuning the general model, thereby improving the model evolution effect of the general pre-trained model.

[0055] (2) An “auxiliary task guidance framework” was proposed, which introduced three auxiliary tasks: “lane line segmentation”, “road marking segmentation”, and “vehicle segmentation”. A multi-task training framework was designed to train these auxiliary tasks together with the target task, thereby helping the model learn the traffic scene-related knowledge required for the target task.

[0056] In order to have a further understanding of the present invention, the following Figures 1 to 3 The present invention describes in detail the construction and efficient evolution method of the visual perception model for UAV aerial images.

[0057] A method for constructing and efficiently evolving a visual perception model for UAV aerial images, including the following steps:

[0058] S1. Dataset preparation: complete dataset selection, data preprocessing and data partitioning.

[0059] 1.1 This example selects the UAVid dataset as the drivable area segmentation dataset for UAV aerial images to verify the invention.

[0060] 1.2 Data preprocessing primarily involves labeling and image enhancement. Labeling involves retaining only the Road label in the UAV ID and assigning all other labels to the background label. Image enhancement includes resizing, mosaic enhancement, random cropping, and random flipping. In this example, we select two enhancement methods: resizing the image to 384×640 pixels and mosaic enhancement with a 50% probability.

[0061] 1.3 Data partitioning is based on the criteria given for the selected dataset. For the UAVid dataset, 200 images are used for training, 70 images for validation, and 150 images for testing. Since the test set labels are not publicly available, this example evaluates the model on the validation set.

[0062] S2. Build a lightweight fine-tuning network.

[0063] The 2.1SAM model is a smaller-scale SAM-B: it contains 12 Transformer layers and the dimension size of the image block embedding is 768.

[0064] 2.2 The hyperparameter settings of the Adapter network are consistent with those of ViT-Adapter-B.

[0065] 2.3 When the VPT network is introduced, 50 trainable tokens are inserted into each layer of the SAM image encoder, and the dimension of the trainable tokens is the same as the embedding dimension of SAM.

[0066] S3. Build an auxiliary task guidance framework. Figure 3As shown in Figure 1, the auxiliary task head is built on the basis of the main task head UPerNet. The specific framework design is as described above and will not be repeated here.

[0067] S4. Train the segmentation network. Use cross-entropy loss to constrain model training. The specific constraint design is as described above and will not be repeated here.

[0068] The back-propagation algorithm is used to update and optimize the network parameter weights until the model loss area converges.

[0069] In this example, image segmentation model training and evaluation are implemented using PyTorch and the MMSegmentation framework. All experiments were conducted on an NVIDIA GeForce RTX 3090 GPU. The batch size was set to 2, and the number of iterations was 80k. The optimizer used was AdamW, with an initial learning rate of 6×10 -5 , the weight decay is 0.01, and the weighting parameters β1 and β2 are set to 0.9 and 0.999 respectively. The learning rate is warmed up in the first 1500 iterations in a linear manner with an increase rate of 1×10 -6 .

[0070] S5. After model training is complete, perform drivable area segmentation detection. Preprocess the drone aerial image to be segmented according to the preprocessing method in S1, then input the preprocessed image into the trained segmentation network. The model output is the drivable area segmentation result.

[0071] After testing, the aAcc, mAcc, and mIoU of the UAVid dataset are 97.41%, 92.73%, and 87.24%, respectively. These results show that the proposed method can effectively complete the task of drivable area segmentation.

[0072] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for constructing and efficiently evolving a visual perception model for drone aerial images, characterized by: The method for constructing and efficiently evolving a visual perception model for drone aerial images includes: S1: Obtain the drivable area segmentation dataset from the drone aerial perspective, complete data preprocessing, and perform data segmentation; S2: Constructing a lightweight fine-tuning network: The lightweight fine-tuning network consists of two parts: the Adapter network and the VPT network. The Adapter network includes the SPM module, the Injector module, and the Extractor module. The SPM module is used to generate multi-scale feature maps, and the Injector module and the Extractor module are used to interact with the generated multi-scale feature maps and the features in the SAM image encoder. The VPT network is used to add trainable tokens to the input of each layer of the Transformer model to indirectly adjust the attention mechanism. S3: Build an auxiliary task guidance framework: Add multiple lightweight auxiliary decoders to the original "encoder-decoder" architecture to perform the "vehicle segmentation" auxiliary task; during training, multiple decoders share a common encoder and perform decoding separately; finally, the losses of these decoders are weighted summed to obtain the final loss; S4: Train the segmentation network: Using the loss function set in S3 and the back-propagation algorithm, iteratively update and optimize the network parameters until the model loss converges; S5: Segment the drivable area: pre-process the drone aerial image to be segmented according to the pre-processing method in S1, and then input the pre-processed image into the trained segmentation network. The output result of the segmentation network model is the drivable area segmentation result.

2. The method for constructing and efficiently evolving a visual perception model for drone aerial images according to claim 1 is characterized in that: Data preprocessing includes label processing and image enhancement; label processing means retaining only the Road label in the UAVid and classifying the remaining labels as background labels; image enhancement includes resizing, mosaic enhancement, random cropping, and random flipping.

3. The method for constructing and efficiently evolving a visual perception model for drone aerial images according to claim 1 is characterized in that: The lightweight fine-tuning network is built on top of the SAM image encoder, and the decoder is replaced with the UPerNet semantic segmentation task head.

4. The method for constructing and efficiently evolving a visual perception model for drone aerial images according to claim 1 is characterized in that: The SPM module adopts a convolutional neural network structure, which includes a Stem network and three convolution layers to obtain four downsampling scales. The following feature map; The three multi-scale feature maps are flattened and concatenated to obtain the feature Input into the Injector module and the Extractor module; among them, represents the i-th multi-scale feature containing spatial prior information, Represents a real-valued feature space of a specific dimension, D represents the number of channels of the feature, and HW represents the length-width product of the image.

5. The method for constructing and efficiently evolving a visual perception model for drone aerial images according to claim 4 is characterized in that: The Injector module uses the cross attention mechanism to integrate multi-scale features The spatial prior information contained in is injected into the Transformer Block corresponding to ViT.

6. The method for constructing and efficiently evolving a visual perception model for drone aerial images according to claim 5 is characterized in that: For the i-th Injector module, the input of the i-th Block As Query, multi-scale features As keys and values, attention calculation is performed according to Realize the injection of spatial prior information, where is the input of the i-th Block, is the input of the updated fusion spatial prior information, γ i is a learnable vector used to balance and attention calculation results.

7. The method for constructing and efficiently evolving a visual perception model for drone aerial images according to claim 6 is characterized in that: The Extractor module consists of a cross-attention and feed-forward neural network, utilizing the previous multi-scale features And the output of the corresponding Transformer Block in ViT Generate new multi-scale features 8. The method for constructing and efficiently evolving a visual perception model for drone aerial images according to claim 7 is characterized in that: according to generate in, To fuse multi-scale features and the output of the i-th Transformer Block of ViT The new features obtained, It is the multi-scale feature output by the i-th Extractor.

9. The method for constructing and efficiently evolving a visual perception model for drone aerial images according to claim 1, characterized in that: The 12 layers in the SAM image encoder are divided into 4 groups, each of which interacts with a pair of Injector-Extractor.

10. The method for constructing and efficiently evolving a visual perception model for drone aerial images according to claim 1, characterized in that: The auxiliary task head is built on the basis of the main task head UPerNet network.

Images