</> Code with Papers by Syntology

Papers › Recursively Refined R-CNN: Instance Segmentation with Self-RoI Rebalancing

Recursively Refined R-CNN: Instance Segmentation with Self-RoI Rebalancing

3 Apr 2021arXiv:2104.01329archive 2025-07-28

Leonardo Rossi, Akbar Karimi, Andrea Prati

Within the field of instance segmentation, most of the state-of-the-art deep learning networks rely nowadays on cascade architectures, where multiple object detectors are trained sequentially, re-sampling the ground truth at each step. This offers a solution to the problem of exponentially vanishing positive samples. However, it also translates into an increase in network complexity in terms of the number of parameters. To address this issue, we propose Recursively Refined R-CNN (R^3-CNN) which avoids duplicates by introducing a loop mechanism instead. At the same time, it achieves a quality boost using a recursive re-sampling technique, where a specific IoU quality is utilized in each recursion to eventually equally cover the positive spectrum. Our experiments highlight the specific encoding of the loop mechanism in the weights, requiring its usage at inference time. The R^3-CNN architecture is able to surpass the recently proposed HTC model, while reducing the number of parameters significantly. Experiments on COCO minival 2017 dataset show performance boost independently from the utilized baseline model. The code is available online at https://github.com/IMPLabUniPr/mmdetection/tree/r3_cnn.

PaperPDFCode

Code

IMPLabUniPr/mmdetection officialmentioned in paperpytorch 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

Instance SegmentationObject DetectionSemantic Segmentation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, DCN) AP50 61.3 #72 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, DCN) AP75 44 #72 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, DCN) APL 56.1 #72 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, DCN) APM 43.6 #72 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, DCN) APS 22.3 #72 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, DCN) mask AP 40.4 #72 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GC-Net) AP50 61.1 #75 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GC-Net) AP75 43.5 #75 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GC-Net) APM 42.8 #75 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GC-Net) APS 22.6 #75 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GC-Net) mask AP 40.2 #75 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GRoIE) AP50 58.8 #77 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GRoIE) AP75 42.3 #77 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GRoIE) APL 54.3 #77 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GRoIE) APM 42.1 #77 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GRoIE) APS 20.7 #77 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN, GRoIE) mask AP 39.1 #77 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN) AP50 58 #80 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN) AP75 41.4 #80 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN) APL 52.8 #80 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN) APM 41 #80 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN) APS 20.4 #80 of 93 Archive leaderboard report
Instance Segmentation COCO minival R3-CNN (ResNet-50-FPN) mask AP 38.2 #80 of 93 Archive leaderboard report
Instance Segmentation coco minval R3-CNN (ResNet-50-FPN, GC-Net) APL 56 #1 of 1 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, DCN) AP50 64.3 #122 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, DCN) AP75 48.9 #122 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, DCN) APL 59.6 #122 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, DCN) APM 48.3 #122 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, DCN) APS 26.6 #122 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, DCN) box AP 44.8 #122 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, GC-Net) AP50 64.1 #133 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, GC-Net) AP75 48.4 #133 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, GC-Net) APL 58.9 #133 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, GC-Net) APM 47.1 #133 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, GC-Net) APS 27 #133 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, GC-Net) box AP 44.3 #133 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN) AP50 61 #159 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN) AP75 46.3 #159 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN) APL 55.7 #159 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN) APM 45.2 #159 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN) APS 24.5 #159 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN) box AP 42 #159 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, GRoIE) AP50 61.2 #217 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, GRoIE) AP75 45.6 #217 of 220 Archive leaderboard report
Object Detection COCO minival R3-CNN (ResNet-50-FPN, GRoIE) APS 24.4 #217 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 ConvolutionConvolutionGRoIENon-Local BlockNon-Local OperationResidual Connection

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