Research and Implementation of Remote Sensing Image Object Detection Algorithm Based on Improved ExtremeNet

According to the characteristics of rich edge information of remote sensing images, DOTA dataset is annotated with edge keypoints on the objects to adapt to the object detection algorithm ExtremeNet. Aiming at the problem of large differences in object scales in remote sensing images, the depthwise separable convolutions and CBAM attention mechanism are introduced to improve ExtremeNet. Experimental results show that the average detection accuracy of the improved ExtremeNet is better.