Single-Shot Object Detection with Enriched Semantics
发布时间
阅读量：
阅读量
Detection incorporating Enhanced Semantic Features for Object Detection and consisting of three components
- one-shot detection model and ssd architecture,
- a segmentation module designed to enhance semantic details at the low-level detection stage
- The segmentation module is utilized to improve the semantic depth of the low-level feature maps obtained from object detectors.
- The segmentation mechanism is employed to enhance the semantic depth of the low-level feature maps obtained from object detectors.
The global activation module enhances semantics in higher-order detection layers. The global activation module uncovers the semantic connections between detection feature channels and object classes in generating higher-level object detection features.
Results：小模型上提升比较明显
全部评论 (0)
还没有任何评论哟~