Papers › Scale-Aware Trident Networks for Object Detection

Scale-Aware Trident Networks for Object Detection

7 Jan 2019ICCV 2019 10arXiv:1901.01892archive 2025-07-28

Yanghao Li, Yuntao Chen, Naiyan Wang, Zhao-Xiang Zhang

Scale variation is one of the key challenges in object detection. In this work, we first present a controlled experiment to investigate the effect of receptive fields for scale variation in object detection. Based on the findings from the exploration experiments, we propose a novel Trident Network (TridentNet) aiming to generate scale-specific feature maps with a uniform representational power. We construct a parallel multi-branch architecture in which each branch shares the same transformation parameters but with different receptive fields. Then, we adopt a scale-aware training scheme to specialize each branch by sampling object instances of proper scales for training. As a bonus, a fast approximation version of TridentNet could achieve significant improvements without any additional parameters and computational cost compared with the vanilla detector. On the COCO dataset, our TridentNet with ResNet-101 backbone achieves state-of-the-art single-model results of 48.4 mAP. Codes are available at https://git.io/fj5vR.

PaperPDFConference PDFCode

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

Code

chengzhengxin/groupsoftmax-simpledet mentioned on GitHubmxnetApache-2.0 report
facebookresearch/detectron2 mentioned on GitHubpytorch report
tusimple/simpledet mentioned on GitHubmxnetApache-2.0 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

ObjectObject Detectionobject-detection

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Object Detection COCO minival TridentNet (ResNet-101) AP50 63.5 #158 of 220 Archive leaderboard report
Object Detection COCO minival TridentNet (ResNet-101) AP75 45.5 #158 of 220 Archive leaderboard report
Object Detection COCO minival TridentNet (ResNet-101) APL 56.9 #158 of 220 Archive leaderboard report
Object Detection COCO minival TridentNet (ResNet-101) APM 47 #158 of 220 Archive leaderboard report
Object Detection COCO minival TridentNet (ResNet-101) APS 24.9 #158 of 220 Archive leaderboard report
Object Detection COCO minival TridentNet (ResNet-101) box AP 42 #158 of 220 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101-Deformable, Image Pyramid) AP50 69.7 #105 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101-Deformable, Image Pyramid) AP75 53.5 #105 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101-Deformable, Image Pyramid) APL 60.3 #105 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101-Deformable, Image Pyramid) APM 51.3 #105 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101-Deformable, Image Pyramid) APS 31.8 #105 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101-Deformable, Image Pyramid) box mAP 48.4 #105 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101) AP50 63.6 #170 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101) AP75 46.5 #170 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101) APL 56.6 #170 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101) APM 46.6 #170 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101) APS 23.9 #170 of 225 Archive leaderboard report
Object Detection COCO test-dev TridentNet (ResNet-101) box mAP 42.7 #170 of 225 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

Introduced by this paper: TridentNet, TridentNet Block

1x1 ConvolutionAverage PoolingBatch NormalizationBottleneck Residual BlockConvolutionDeformable ConvolutionDilated ConvolutionGlobal Average PoolingKaiming InitializationMax PoolingRandom Horizontal FlipReLUResidual BlockResidual ConnectionSoft-NMSStep DecayTridentNetTridentNet Block

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