Research map / 3D vision / 3D generation
SeeThrough3D: Occlusion Aware 3D Control in Text-to-Image Generation
Vaibhav Agrawal, Rishubh Parihar, Pradhaan S Bhat, Ravi Kiran Sarvadevabhatla, Venkatesh Babu Radhakrishnan
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 |
|---|---|---|---|
| render boxes translucent instead of opaque (ablation) | OSCR representation · 3DOcBench | depth ordering: 1.20 -> 1.46 | Improved |
| render boxes translucent instead of opaque (ablation) | OSCR representation · 3DOcBench | angular error: 46.15 -> 47.92 | Hurt |
| add color-coded box faces (ablation) | OSCR representation · 3DOcBench | angular error: 88.77 -> 47.92 | Improved |
| add color-coded box faces (ablation) | OSCR representation · 3DOcBench | depth ordering: 1.36 -> 1.46 | Improved |
| add attention-masking object binding (ablation) | SeeThrough3D · 3DOcBench | CLIP objectness score: 20.45 -> 22.86 | Improved |
| add attention-masking object binding (ablation) | SeeThrough3D · 3DOcBench | depth ordering: 0.98 -> 1.46 | Improved |
| filter out overly simplistic low-occlusion training layouts (ablation) | SeeThrough3D training data · 3DOcBench | depth ordering: 1.24 -> 1.46 | Improved |
| filter out overly simplistic low-occlusion training layouts (ablation) | SeeThrough3D training data · 3DOcBench | KID (x10^-3): 6.34 -> 5.43 | Improved |
| replace 3D layout depth maps (LooseControl) with OSCR translucent box rendering | 3D layout conditioning · 3DOcBench | angular error: 89.88 -> 47.92 | Improved |
| replace 2D object layers (LaRender) with 3D-aware OSCR | occlusion control · 3DOcBench | depth ordering: 1.02 -> 1.46 | Improved |
| use sequential generation-inversion cycles (Build-A-Scene) instead of single-pass conditioning | 3D layout control · 3DOcBench | KID (x10^-3): 5.43 -> 20.12 | Hurt |
Closest papers by meaning
- Build-A-Scene: Interactive 3D Layout Control for Diffusion-Based Image Generation (ICLR 2025 · similarity 0.88)
- Sculpt3D: Multi-View Consistent Text-to-3D Generation with Sparse 3D Prior (CVPR 2024 · similarity 0.87)
- Slice3D: Multi-Slice Occlusion-Revealing Single View 3D Reconstruction (CVPR 2024 · similarity 0.86)
- Layout-your-3D: Controllable and Precise 3D Generation with 2D Blueprint (ICLR 2025 · similarity 0.86)
- 3D Occupancy Prediction with Low-Resolution Queries via Prototype-aware View Transformation (CVPR 2025 · similarity 0.86)
- Cue3D: Quantifying the Role of Image Cues in Single-Image 3D Generation (NeurIPS 2025 · 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.
