Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• The AffordBridge dataset contains 291,637 functional interaction annotations across 685 high-resolution indoor scenes.result 95%
• The AffordMatcher method outperforms the second-best baseline by 7.8 in overall mAP for functionality affordance segmentation.result 90%
• AffordMatcher establishes semantic correspondences between image-based and point cloud-based instances for keypoint matching to identify affordance regions.method 85%
• The AffordMatcher model contains 20.7 million parameters, which is fewer than some baseline models like Ego-SAG (24.8M).technical 90%
• Existing datasets for 3D scene affordance learning (e.g., Scenefun3D, MIPA) have fewer samples and limited affordance diversity compared to AffordBridge.limitation 80%
• AffordMatcher generates more compact and accurate affordance masks for fine-grained interaction parts (like knobs and prongs) compared to baselines like Ego-SAG.result 75%
• The dataset construction pipeline uses human pose estimation (OpenPose) to annotate keypoint-based hand-object contacts to address ambiguities in static visual signifiers.method 85%
• AffordMatcher achieves an mAP@0.50 of 69.7, which is higher than all listed baselines.result 95%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card