Research map / 3D vision / Depth and stereo
SEDiff: Structure Extraction for Domain Adaptive Depth Estimation via Denoising Diffusion Models
Dongseok Shim, Hyoun Jin Kim
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 |
|---|---|---|---|
| feed extracted structural content instead of raw RGB to the depth network (ablation) | SEDiff · KITTI Eigen test (80m) | Abs Rel: 0.268 -> 0.165 | Improved |
| convert input to grayscale before the geometric feature extractor (ablation) | SEDiff · KITTI Eigen test (80m) | Abs Rel: 0.268 -> 0.231 | Improved |
| train G with diffusion instead of using the LDM autoencoder encoder E (ablation) | SEDiff · KITTI Eigen test (80m) | Abs Rel: 0.183 -> 0.165 | Improved |
| latent-level instead of pixel-level diffusion (ablation) | SEDiff · KITTI Eigen test (80m) | Abs Rel: 0.166 -> 0.165 | No clear effect |
| estimate attention map from style-transferred image rather than raw synthetic image (ablation) | SEDiff · KITTI Eigen test (80m) | Abs Rel: 0.173 -> 0.168 | Improved |
| ensemble depth from both synthetic and style-transferred inputs (ablation) | SEDiff · KITTI Eigen test (80m) | Abs Rel: 0.168 -> 0.165 | Improved |
| SEDiff vs 3D-PL on KITTI with monocular real images | SEDiff · KITTI Eigen test (80m) | Abs Rel / Sq Rel: Abs Rel 0.169 -> 0.165; Sq Rel 1.262 -> 1.301 | Mixed |
| add real stereo pairs to training | SEDiff · KITTI Eigen test (80m) | Abs Rel: 0.165 -> 0.110 | Improved |
| SEDiff vs 3D-PL on unseen DrivingStereo | SEDiff · DrivingStereo (unseen) | Abs Rel: 0.243 -> 0.193 | 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.
- SharinGAN — SharinGAN: Combining Synthetic and Real Data for Unsupervised Geometry Estimation (CVPR 2020) · better on 2 of 2 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
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · similarity 0.88)
- SharpDepth: Sharpening Metric Depth Predictions Using Diffusion Distillation (CVPR 2025 · similarity 0.88)
- DCDepth: Progressive Monocular Depth Estimation in Discrete Cosine Domain (NeurIPS 2024 · similarity 0.88)
- Repurposing Diffusion-Based Image Generators for Monocular Depth Estimation (CVPR 2024 · similarity 0.87)
- Pixel-Perfect Depth with Semantics-Prompted Diffusion Transformers (NeurIPS 2025 · similarity 0.87)
- Depth Anything V2 (NeurIPS 2024 · similarity 0.86)
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.
