FA-SSD: A Small Object Detection Algorithm Based on Feature Alignment
Yanwen Zheng, Yu Wang, Fan Li, Yingmei Zhu · 2022 5th International Conference on Pattern Recognition and Artificial Intelligence (PRAI) · 2022
In order to solve the problem that the SSD algorithm has low detection accuracy for small objects, the FA-SSD algorithm is proposed. The FA-SSD algorithm improves the upsampling path, enhances the semantics of the shallow feature map, and then combines the feature alignment and SE module to design a feature fusion module to fuse features more effectively. The experimental results show that the MAP of the FA-SSD algorithm on the VOC07-test dataset reaches 79.81%, which is 2.01% higher than that of the SSD algorithm, and in our small target dataset, the MAP is 3.61% higher than that of the SSD algorithm, and the FA-SSD algorithm has better detection performance.