AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

X-Pose: Detecting Any Keypoints

Jie Yang, Ailing Zeng, Ruimao Zhang, Lei Zhang

ECCV 2024 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary 3D and world perception · 45 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add object-level alignment contrastive loss (ablation)X-Pose · AP-10Kkeypoint-level AP: 45.5 -> 72.6Improved
add keypoint-level alignment contrastive loss (ablation)X-Pose · AP-10Kkeypoint-level AP: 72.6 -> 73.2Improved
train with both visual and textual prompts rather than visual only (ablation)X-Pose · AP-10K (visual test prompt)keypoint-level AP: 71.5 -> 72.8Improved
train with both prompts rather than textual only (ablation)X-Pose · AP-10K (textual test prompt)keypoint-level AP: 72.9 -> 73.2Improved
add 5 animal datasets to training (ablation)X-Pose · AnimalPose (unseen)PCK: 52.7 -> 70.1Improved
add part-level, rigid and soft object datasets (ablation)X-Pose · AP-10K keypoint APAP: AP-10K 74.2 -> 74.0; AnimalPose 70.1 -> 73.4Mixed
fine-tune GroundingDINO on keypoint datasetsGroundingDINO-T · COCO valkeypoint-level AP: 3.1 -> 1.8Hurt
use end-to-end X-Pose instead of top-down CLAMP in zero-shot settingtextual prompt keypoint detection · AP-10K train Bovidae test CanidaeAP: 46.9 -> 53.8Improved
use X-Pose instead of top-down SOTA on COCOgeneral keypoint detection · COCO valAP: 78.6 -> 76.8Hurt
X-Pose versus CLIP for text-to-keypoint alignmenttext-to-image alignment · AP-10K val keypointCLIP score: 21.75 -> 66.01Improved

Later papers that compared with this method

1 later accepted paper reports X-Pose as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy