Research map / 3D vision / Depth and stereo
Real2Sim2Real: RetinalDepth-64K for Depth Estimation in Posterior Segment Ophthalmic Surgery
Bingwen Dong, Gan Liu, Xiaoxi Lu, Guangcheng Chen, Jialu Zhang, Yan Hu, Xiaoqing Zhang, Jiang Liu
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 |
|---|---|---|---|
| fine-tune Depth Anything on RetinalDepth | Depth Anything · RetinalDepth single-image test (whole scene) | d1: 0.591 -> 0.975 | Improved |
| fine-tune VGGT on RetinalDepth | VGGT · RetinalDepth single-image test (whole scene) | Abs Rel: 0.322 -> 0.061 | Improved |
| fine-tune Marigold on RetinalDepth | Marigold · RetinalDepth single-image test (whole scene) | d1: 0.587 -> 0.885 | Improved |
| fine-tune DUSt3R on RetinalDepth | DUSt3R · RetinalDepth single-image test (whole scene) | d1: 0.015 -> 0.872 | Improved |
| fine-tune MoGE on RetinalDepth | MoGE · RetinalDepth single-image test (whole scene) | d1: 0.430 -> 0.935 | Improved |
| fine-tune stereo models and evaluate on instrument region | VGGT · RetinalDepth single-image test (instrument region) | d1: 0.612 (whole, zero-shot) -> 0.339 (instrument, fine-tuned) | Mixed |
| fine-tune monocular models and evaluate on instrument region | Depth Anything V2 · RetinalDepth single-image test (instrument region) | d1: 0.889 | Improved |
| use stereo VGGT instead of monocular models zero-shot | VGGT · RetinalDepth single-image test (zero-shot) | d1: 0.591 -> 0.612 | Improved |
| use fine-tuned single-image models frame-by-frame instead of zero-shot video models | VGGT · RetinalDepth video test | d1: 0.228 -> 0.956 | Improved |
| use video-specific model VDA-L instead of fine-tuned VGGT for temporal stability | VDA-L · RetinalDepth video test | TDV: 5.870 -> 0.044 | Improved |
| use frame-by-frame stereo VGGT for temporal consistency | VGGT · RetinalDepth video test | TDV: 5.870 | Hurt |
Closest papers by meaning
- FOCUS: Benchmarking Retinal Model Generalization from Foundation Vision Encoders to Multimodal LLMs (NeurIPS 2026 · similarity 0.85)
- Towards Dynamic 3D Reconstruction of Hand-Instrument Interaction in Ophthalmic Surgery (NeurIPS 2025 · similarity 0.84)
- OphCLIP: Hierarchical Retrieval-Augmented Learning for Ophthalmic Surgical Video-Language Pretraining (ICCV 2025 · similarity 0.84)
- RealX3D: A Physically-Degraded 3D Benchmark for Multi-view Visual Restoration and Reconstruction (IJCV 2026 · similarity 0.83)
- Efficient Depth Estimation for Unstable Stereo Camera Systems on AR Glasses (CVPR 2025 · similarity 0.83)
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · similarity 0.82)
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.
