Dense Point-to-Mask Optimization with Reinforced Point Selection for Crowd Instance Segmentation

Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• Dense Point-to-Mask Optimization (DPMO) integrates SAM with Nearest Neighbor Exclusive Circle (NNEC) constraint to generate dense instance segmentation from point annotations.method 95%
• The proposed method achieves state-of-the-art crowd instance segmentation performance on ShanghaiTech, UCF-QNRF, JHU-CROWD++, and NWPU-Crowd datasets.result 92%
• Directly applying currently popular large foundation models such as SAM does not yield ideal results in dense crowds.limitation 88%
• Mask annotations obtained through segmentation help improve the accuracy of region labels and resolve correspondence between individual location coordinates and crowd density maps.technical 85%
• The Reinforced Point Selection (RPS) framework trained with Group Relative Policy Optimization (GRPO) selects the best predicted point from a sampling of the initial point prediction.method 93%
• New loss functions supervised by masks boost counting performance across different models.result 90%
• Mask annotations play a significant role in enhancing counting accuracy.technical 87%
• Point labels are common in crowd datasets while region labels are rare and inaccurate.limitation 82%

Competitive landscape

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

Startup potential card