EdgeYOLO: An Edge-Real-Time Object Detector
Shihan Liu, Junlin Zha, Jian Sun, Zhuo Li, Gang Wang
Abstract
An efficient, low-complexity, and anchor-free object detector based on the state-of-the-art YOLO framework is proposed in this paper, which can be implemented in real time on edge computing platforms. An enhanced data augmentation method is developed to effectively suppress overfitting during training, and a hybrid random loss function is designed to improve the detection accuracy of small objects. Inspired by FCOS, a lighter and more efficient decoupled head is proposed, and its inference speed can be improved with little loss of precision. Our baseline model can reach the accuracy of 50.6% AP50:95 and 69.8% AP <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">50</inf> in MS COC02017 dataset, 26.9% AP <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">50:95</inf> and 45.4% AP <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">50</inf> in VisDrone2019-DET dataset, and it meets real-time requirements (FPS230) on edge-computing device Nvidia Jetson AGX Xavier. And as is shown in Fig. 1, lighter models with less parameters designed for edge computing devices with lower computing power also show better performances. Our source code, hyper-parameters and model weights are all available at https://github.com/LSH9832/edgeyolo.