SS-YOLO: An Object Detection Algorithm based on YOLOv3 and ShuffleNet

Supported by the improvement in the computing power of devices and increasing amount of data is being produced by digital society, the object detection framework based on deep learning has developed rapidly. Based on R-CNN, SSD, YOLO and other classical frameworks, many excellent object detection frameworks have appeared. YOLOv3 is at the current leading level in various aspects such as detection speed and accuracy. YOLOv3-Tiny can achieve real-time detection on the common GPU, but the accuracy is lower. SS-YOLO builds a feature extraction network with reference to the structure of ShuffleNet, which improves accuracy while ensuring the speed. The idea of SENet is added to the structure to adjust the degree of attention to the channel according to the importance of the characteristics of different channels. Compared to YOLOv3-Tiny, SS-YOLO improves the mean average precision by 6.5% and frames per second by 4 on the PASCAL VOC2007 dataset.