Papers › Sparse R-CNN: End-to-End Object Detection with Learnable Proposals

Sparse R-CNN: End-to-End Object Detection with Learnable Proposals

25 Nov 2020CVPR 2021 1arXiv:2011.12450archive 2025-07-28

Peize Sun, Rufeng Zhang, Yi Jiang, Tao Kong, Chenfeng Xu, Wei Zhan, Masayoshi Tomizuka, Lei LI, Zehuan Yuan, Changhu Wang, Ping Luo

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×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× 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.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

PeizeSun/SparseR-CNN officialmentioned in papermentioned on GitHubpytorchMIT report
Booomshaker/SparseRCNNWSL mentioned on GitHubpytorch report
henbucuoshanghai/sparsercnn mentioned on GitHubpytorch report
liangheming/sparse_rcnnv1 mentioned on GitHubpytorch report
open-mmlab/mmdetection pytorchApache-2.0 report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

2D Object DetectionObjectObject DetectionObject Recognitionobject-detection

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
2D Object Detection CeyMo Sparse R-CNN mAP 47.3 #5 of 5 Archive leaderboard report
2D Object Detection SARDet-100K Sparse R-CNN box mAP 38.1 #12 of 13 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, learnable proposals, random crop aug, FPN) AP50 64.6 #112 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, learnable proposals, random crop aug, FPN) AP75 49.5 #112 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, learnable proposals, random crop aug, FPN) APL 61.6 #112 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, learnable proposals, random crop aug, FPN) APM 48.3 #112 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, learnable proposals, random crop aug, FPN) APS 28.3 #112 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, learnable proposals, random crop aug, FPN) box AP 45.6 #112 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, learnable proposals, random crop aug, FPN) AP50 63.4 #128 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, learnable proposals, random crop aug, FPN) AP75 48.2 #128 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, learnable proposals, random crop aug, FPN) APL 59.5 #128 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, learnable proposals, random crop aug, FPN) APM 47.2 #128 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, learnable proposals, random crop aug, FPN) APS 26.9 #128 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, learnable proposals, random crop aug, FPN) box AP 44.5 #128 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, FPN) AP50 62.1 #138 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, FPN) AP75 47.2 #138 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, FPN) APL 59.7 #138 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, FPN) APM 46.3 #138 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, FPN) APS 26.1 #138 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-101, FPN) box AP 43.5 #138 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, FPN) AP50 61.2 #155 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, FPN) AP75 45.7 #155 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, FPN) APL 57.6 #155 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, FPN) APM 44.6 #155 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, FPN) APS 26.7 #155 of 220 Archive leaderboard report
Object Detection COCO minival Sparse R-CNN (ResNet-50, FPN) box AP 42.3 #155 of 220 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

1x1 ConvolutionConvolutionFPNSparse R-CNN

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections