Sparse R-CNN: End-to-End Object Detection with Learnable Proposals
This work simplifies the object detection pipeline for researchers and practitioners by removing the complex design and assignment of dense object candidates and the NMS post-processing step.
This paper introduces Sparse R-CNN, an object detection method that replaces dense, hand-designed object candidates with a fixed, sparse set of 100 learnable proposals. This approach eliminates the need for non-maximum suppression and achieves 45.0 AP on the COCO dataset with a ResNet-50 FPN model, running at 22 fps.
We present Sparse R-CNN, a purely sparse method for object detection in images. Existing works on object detection heavily rely on dense object candidates, such as $k$ anchor boxes pre-defined on all grids of image feature map of size $H\times W$. In our method, however, a fixed sparse set of learned object proposals, total length of $N$, are provided to object recognition head to perform classification and location. By eliminating $HWk$ (up to hundreds of thousands) hand-designed object candidates to $N$ (e.g. 100) learnable proposals, Sparse R-CNN completely avoids all efforts related to object candidates design and many-to-one label assignment. More importantly, final predictions are directly output without non-maximum suppression post-procedure. Sparse R-CNN demonstrates accuracy, run-time and training convergence performance on par with the well-established detector baselines on the challenging COCO dataset, e.g., achieving 45.0 AP in standard $3\times$ training schedule and running at 22 fps using ResNet-50 FPN model. We hope our work could inspire re-thinking the convention of dense prior in object detectors. The code is available at: https://github.com/PeizeSun/SparseR-CNN.