Object detection method and system based on multi-scale feature map reconstruction and knowledge distillation

Abstract

The invention discloses a target detection method and system based on multi-scale feature map reconstruction and knowledge distillation. The method first uses the backbone network Darknet-53 to extract features, and the deep features generate multi-scale features through upsampling and shallow feature tensor splicing. Feature map; then use the feature re-calibration strategy to automatically obtain the weight of each channel in the feature map, promote useful features and suppress useless features according to the weight, and then use the residual module to fuse the semantic information of the top-level features and the details of the underlying features; Then, the γ coefficient of the batch normalization layer in the backbone network is introduced into the pruning objective function for training, and the channel where the γ coefficient below the threshold is located is removed from the model according to the pruning threshold; finally, the trained YOLOv3 benchmark model is used as the teacher. network, the pruned model is used as a student network for knowledge distillation. The invention improves the accuracy of detecting objects of different sizes in a large range, reduces the calculation amount of the model, and improves the model detection speed.

Description

technical field [0001] The invention provides a target detection method and system based on multi-scale feature map reconstruction and knowledge distillation, belonging to the technical field of target detection of computer vision. Background technique [0002] Image target recognition is a research topic involving computer vision, pattern recognition and artificial intelligence. With the rapid development of hardware technology, embedded intelligent devices based on deep learning platforms are becoming more and more mature, and more and more detection algorithms are embedded However, the traditional detection method has a large difference in detection accuracy for targets of different sizes within a certain range, cannot accurately identify the target, and cannot meet daily needs, and the traditional detection algorithm model has too many parameters, which requires more computing power. Therefore, it is necessary to propose a technology that can not only make the detection ...

AI Technical Summary

Problems solved by technology

Low-resolution, semantically strong features are upsampled and combined with high-resolution, semantically weak features to build a feature pyramid that shares rich semantics at all levels, but it still has a lot of room for improvement, For example, [Shu Liu, Lu Qi, Haifang Qin, Jianping Shi, and Jiaya Jia. Path aggregation network for instance segmentation. In CVPR, 2018] manually designed the fusion structure and enhanced feature fusion to improve the detection accuracy a lot, but these algorithms It has not been optimized in combination with YOLOv3 and the actual scene, and there is still a lot of room for improvement in feature map reconstruction

[0005] For the model compression method of target detection, many previous works have been proposed to compress large CNNs or directly learn more effective CNN models for fast reasoning, such as literature [E.L.Denton, W.Zaremba, J.Bruna, Y.LeCun, and R.Fergus.Exploiting linear structure within convolutional networks efficient evaluation.In NIPS,2014.] Applied low-rank approximation, literature [S.Han,J. Pool,J.Tran,and W.Dally.Learning both weights and connections for efficient neural network.In NIPS, pages 1135–1143,2015.] The weight pruning used, etc., but most of these techniques require specially designed software / hardware accelerators to accelerate execution, aiming at the target detection model on embedded devices There are relatively few compression methods. The current compression algorithm compresses YOLOv3 with high time complexity, and cannot well complete the target detection tasks in embedded device application scenarios (such as pedestrian and vehicle target detection in intelligent transportation)

Images