A Mixed Two-stage Object Detector for Image Processing of Power System Applications
Xitian Long, Zhe Zheng, Yingying Chi, Rui Liu · 2020
Object detection algorithms based on deep learning has become increasingly important in Image Processing. This paper proposes a deep learning method for improving the objects detection accuracy while supporting a real-time operation by optimizing SSD model into a mixed Two-stage detection model with an extra dense small object detection module which can localize the precise position and give the classification of the dense small objects, using a fully convolutional network. Compared with the SOTA detection model such as two-stage Faster R-CNN and single-stage SSD, our approach has a better trade-off between accuracy and efficiency. The experimental results outperform Faster R-CNN in detection speed and considerably better than SSD in detection accuracy.