Papers › DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

DETRPose: Real-time end-to-end transformer model for multi-person pose estimation

16 Jun 2025arXiv:2506.13027archive 2025-07-28

Sebastian Janampa, Marios Pattichis

Multi-person pose estimation (MPPE) estimates keypoints for all individuals present in an image. MPPE is a fundamental task for several applications in computer vision and virtual reality. Unfortunately, there are currently no transformer-based models that can perform MPPE in real time. The paper presents a family of transformer-based models capable of performing multi-person 2D pose estimation in real-time. Our approach utilizes a modified decoder architecture and keypoint similarity metrics to generate both positive and negative queries, thereby enhancing the quality of the selected queries within the architecture. Compared to state-of-the-art models, our proposed models train much faster, using 5 to 10 times fewer epochs, with competitive inference times without requiring quantization libraries to speed up the model. Furthermore, our proposed models provide competitive results or outperform alternative models, often using significantly fewer parameters.

PaperPDFCode

Code

SebastianJanampa/DETRPose officialmentioned in papermentioned on GitHubpytorch 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 Pose EstimationDecoderMulti-Person Pose EstimationPose EstimationQuantization

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Multi-Person Pose Estimation CrowdPose DETRPose-X AP Easy 81.3 #5 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-X AP Hard 68.1 #5 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-X AP Medium 75.7 #5 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-X mAP @0.5:0.95 75.1 #5 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-L AP Easy 79.5 #6 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-L AP Hard 66.1 #6 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-L AP Medium 74.0 #6 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-L mAP @0.5:0.95 73.3 #6 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-M AP Easy 78.6 #9 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-M AP Hard 64.5 #9 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-M AP Medium 72.6 #9 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-M mAP @0.5:0.95 72.0 #9 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-S AP Easy 74.7 #16 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-S AP Hard 59.3 #16 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-S AP Medium 68.1 #16 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-S mAP @0.5:0.95 67.4 #16 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-N AP Easy 65.0 #26 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-N AP Hard 46,6 #26 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-N AP Medium 56,6 #26 of 28 Archive leaderboard report
Multi-Person Pose Estimation CrowdPose DETRPose-N mAP @0.5:0.95 56.0 #26 of 28 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

SPEED

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