Research map / Video / Video generation
Mitigating Surgical Data Imbalance with Dual-Prediction Video Diffusion Model
Danush Kumar Venkatesh, Adam Schmidt, Muhammad Abdullah Jamal, Omid Mohareri
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add SurgiFlowVid synthetic videos (Seg conditioning) for under-represented classes | MViTv2 action recognition · SAR-RARP50 | mean Jaccard: 0.46 -> 0.52 | Improved |
| add dual-prediction module (RGB plus optical flow) to SurV-Gen (ablation) | SurgiFlowVid · SAR-RARP50 action recognition | mean Jaccard: 0.46 -> 0.49 | Improved |
| add sparse visual encoder with RGB or mask conditioning (ablation) | SurgiFlowVid · SAR-RARP50 action recognition | mean Jaccard: 0.49 -> 0.52 (Seg.) | Improved |
| model each under-represented class individually with RGB conditioning (ablation) | SurgiFlowVid · SAR-RARP50 action recognition | mean Jaccard: 0.46 -> 0.53; A7 0.32 -> 0.52 | Improved |
| model each class individually with segmentation conditioning (ablation) | SurgiFlowVid · SAR-RARP50 action recognition | mean Jaccard: 0.52 -> 0.48 | Mixed |
| duplicate the training set rather than add synthetic samples (ablation) | downstream action recognizer · SAR-RARP50 | mean Jaccard: 0.46 -> 0.45 | No clear effect |
| shuffle frames of synthetic videos (ablation) | downstream action recognizer · SAR-RARP50 | mean Jaccard: 0.46 -> 0.43 | Hurt |
| add noise or use only sparse frames as synthetic data (ablation) | downstream action recognizer · SAR-RARP50 | mean Jaccard: 0.46 -> 0.42 (sparse), 0.43 (noisy) | Hurt |
| add SurgiFlowVid RGB-conditioned samples to real GraSP data | MViTv2 step recognition · GraSP | mean mAP: 0.47 -> 0.52 | Improved |
| add Endora or SparseCtrl (RGB) synthetic samples | MViTv2 step recognition · GraSP | mean mAP: 0.47 -> 0.45 (Endora), 0.46 (SparseCtrl) | Hurt |
| add SurgiFlowVid (Seg) synthetic samples for tool presence detection | tool detector · SAR-RARP50 | mean Dice: 0.69 -> 0.79 | Improved |
| add SurgFlowVid (Seg) synthetic samples for tool detection | tool detector · GraSP | mean Dice: 0.68 -> 0.72 | Improved |
Closest papers by meaning
- SurgVista: Long-Horizon Surgical World Modeling with Plausible Instrument-Tissue Dynamics (NeurIPS 2026 · similarity 0.86)
- CholecTrack20: A Multi-Perspective Tracking Dataset for Surgical Tools (CVPR 2025 · similarity 0.85)
- Bridging Vision and Language for Robust Context-Aware Surgical Point Tracking: The VL-SurgPT Dataset and Benchmark (AAAI 2026 · similarity 0.85)
- Procedure-Aware Surgical Video-language Pretraining with Hierarchical Knowledge Augmentation (NeurIPS 2024 · similarity 0.84)
- SurgCoT: Advancing Spatiotemporal Reasoning in Surgical Videos through a Chain-of-Thought Benchmark (CVPR 2026 · similarity 0.84)
- AsymRnR: Video Diffusion Transformers Acceleration with Asymmetric Reduction and Restoration (ICML 2025 · similarity 0.83)
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.
