Object Detection Made Simpler by Eliminating Heuristic NMS
This work addresses the need for simpler and more efficient object detection pipelines by removing the heuristic NMS step, which is incremental as it builds on existing one-stage detectors.
The paper tackles the problem of eliminating heuristic non-maximum suppression (NMS) in object detection by proposing a simple NMS-free, end-to-end framework that modifies a one-stage detector like FCOS, achieving on-par or improved accuracy on the COCO dataset with similar inference speed.
We show a simple NMS-free, end-to-end object detection framework, of which the network is a minimal modification to a one-stage object detector such as the FCOS detection model [Tian et al. 2019]. We attain on par or even improved detection accuracy compared with the original one-stage detector. It performs detection at almost the same inference speed, while being even simpler in that now the post-processing NMS (non-maximum suppression) is eliminated during inference. If the network is capable of identifying only one positive sample for prediction for each ground-truth object instance in an image, then NMS would become unnecessary. This is made possible by attaching a compact PSS head for automatic selection of the single positive sample for each instance (see Fig. 1). As the learning objective involves both one-to-many and one-to-one label assignments, there is a conflict in the labels of some training examples, making the learning challenging. We show that by employing a stop-gradient operation, we can successfully tackle this issue and train the detector. On the COCO dataset, our simple design achieves superior performance compared to both the FCOS baseline detector with NMS post-processing and the recent end-to-end NMS-free detectors. Our extensive ablation studies justify the rationale of the design choices.