Research map / 3D vision / Depth and stereo
DepthFocus: Controllable Depth Estimation for See-Through Scenes
Junhong Min, Jimin Kim, Minwook Kim, Cheol-Hui Min, Youngpil Jeon, Minyong Choi
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 conditional Mixture-of-Experts routing (ablation) | DepthFocus · Synthetic multi-layer benchmark, transmissive first layer | Bad-2: 13.97 -> 11.78 | Improved |
| add direct condition injection (DCI) (ablation) | DepthFocus · Synthetic multi-layer benchmark, transmissive first layer | Bad-2: 14.75 -> 11.78 | Improved |
| add auxiliary segmentation loss (ablation) | DepthFocus · Synthetic benchmark, transmissive first layer | Bad-2: 13.15 -> 11.78 | Improved |
| add auxiliary segmentation loss (ablation) | DepthFocus · Synthetic benchmark, last layer | Bad-2: 38.16 -> 37.70 without it | No clear effect |
| add data curation using large single-disparity datasets (ablation) | DepthFocus · Synthetic benchmark, transmissive first layer | Bad-2: 15.59 -> 11.78 | Improved |
| replace S2M2 fine-tuning with steerable DepthFocus on same data | S2M2 (ft) · Booster reflective/transmissive | EPE: 7.69 -> 3.41 | Improved |
| use DepthFocus instead of S2M2 on Middlebury v3 | S2M2 · Middlebury v3 non-occlusion | Bad-4: 0.54 -> 0.60 | Hurt |
| use steerable DepthFocus instead of static multi-head baselines | RAFT-4layer/ASGrasp · LayeredFlow val layer 2 | EPE: 15.59 -> 7.77 | Improved |
| use DepthFocus on real acrylic bilayer scenes vs monocular-prior stereo | FoundationStereo · Real bilayer 60% first layer | Bad-4: 99.88 -> 8.72 | Improved |
| use 2-layer fine-tuned ASGrasp vs 4-layer variant | ASGrasp · Synthetic benchmark, transmissive first layer | Bad-2: 57.65 (4layer) -> 45.73 (2layer) | Improved |
Closest papers by meaning
- DepthCues: Evaluating Monocular Depth Perception in Large Vision Models (CVPR 2025 · similarity 0.91)
- DualFocus: Depth from Focus with Spatio-Focal Dual Variational Constraints (NeurIPS 2025 · similarity 0.90)
- Depth2Pose: A Pose-Based Benchmark for Monocular Depth Estimation without Ground-Truth Depth (NeurIPS 2026 · similarity 0.90)
- DepthMaster: Unified Monocular Depth Estimation for Perspective and Panoramic Images (NeurIPS 2026 · similarity 0.89)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · similarity 0.89)
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · similarity 0.89)
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.
