Research map / Video / Video generation
ICDepth: Taming Video Diffusion Models for Video Depth Estimation via In-Context Conditioning
Xuanhua He, JIAXIN XIE, Mingzhe Zheng, Qifeng Chen
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 in-context conditioning instead of channel concatenation (ablation) | ICDepth · Sintel ablation | AbsRel: 0.367 -> 0.250 | Improved |
| add in-context conditioning instead of channel concatenation (ablation) | ICDepth · Sintel ablation | delta1: 0.654 -> 0.749 | Improved |
| add SAND-Attention instead of full attention (ablation) | ICDepth · Sintel ablation | AbsRel: 0.413 -> 0.250 | Improved |
| add SAND-Attention instead of full attention (ablation) | ICDepth · Sintel ablation | delta1: 0.443 -> 0.749 | Improved |
| add RoPE alignment (ablation) | ICDepth · Sintel ablation | AbsRel: 0.410 -> 0.250 | Improved |
| add decoupled attention (ablation) | ICDepth · Sintel ablation | AbsRel: 0.262 -> 0.250 | Improved |
| add decoupled attention (ablation) | ICDepth · Sintel ablation | delta1: 0.710 -> 0.749 | Improved |
| add SRFM module (ablation) | ICDepth · Sintel ablation | AbsRel: 0.306 -> 0.250 | Improved |
| add DINOv2 features (ablation) | ICDepth · Sintel ablation | AbsRel: 0.269 -> 0.250 | Improved |
| add resolution embedding (ablation) | ICDepth · Sintel ablation | AbsRel: 0.264 -> 0.250 | Improved |
| increase sampling steps from 5 to 20 (ablation) | ICDepth · Sintel | AbsRel: 0.250 -> 0.217 | Mixed |
| ICDepth versus Video Depth Anything on Sintel | ICDepth · zero-shot Sintel (50 frames) | AbsRel: 0.383 -> 0.250 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DepthCrafter — DepthCrafter: Generating Consistent Long Depth Sequences for Open-world Videos (CVPR 2025) · better on 5 of 5 shared results
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
- Learning Temporally Consistent Video Depth from Video Diffusion Priors (CVPR 2025 · similarity 0.92)
- Video Depth Anything: Consistent Depth Estimation for Super-Long Videos (CVPR 2025 · similarity 0.89)
- BetterDepth: Plug-and-Play Diffusion Refiner for Zero-Shot Monocular Depth Estimation (NeurIPS 2024 · similarity 0.88)
- Align3R: Aligned Monocular Depth Estimation for Dynamic Videos (CVPR 2025 · similarity 0.88)
- ECoDepth: Effective Conditioning of Diffusion Models for Monocular Depth Estimation (CVPR 2024 · similarity 0.88)
- Depth Any Video with Scalable Synthetic Data (ICLR 2025 · similarity 0.88)
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.
