Animal Object Detection System based on the Improved YOLOX

For the deteriorating living environment of wildlife, it is very important to protect and monitor wildlife. An object detection algorithm based on improved YOLOx is put forward here to settle these problems. The coordinated attention mechanism is introduced into the trunk network of YOLOx. At the same time, the pre-training weight file of large target detection is adopted to embed the position information into the guidance channel, so the network can follow in the more area to locate the detection target more accurately. Finally, experiment on PASCAL VOC data set and the consequences show that the accuracy and the recognition effect of the algorithm improved by 2% compared to YOLOx, which proves the effectiveness of this algorithm.