Research map / 3D vision / Face and shape
Unsupervised Pixel-Level Semantic Left-Right Understanding of In-the-Wild Images
Weikang Wang, Tobias Weißberg, Florian Bernard
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 L_var (total variation loss) (ablation) | Pix2LR · SPair-71k | acc-hat: 55.6 -> 83.8 | Improved |
| add L_dis (dissimilarity loss) (ablation) | Pix2LR · SPair-71k | acc-hat: 56.8 -> 83.8 | Improved |
| add L_fif (fifty-fifty loss) (ablation) | Pix2LR · SPair-71k | acc-hat: 60.9 -> 83.8 | Improved |
| add L_img (image symmetry loss) (ablation) | Pix2LR · SPair-71k | acc-hat: 73.8 -> 83.8 | Improved |
| add L_var (ablation) | Pix2LR · BeCoS rendered | acc-hat: 53.9 -> 87.4 | Improved |
| add L_dis (ablation) | Pix2LR · BeCoS rendered | acc-hat: 50.3 -> 87.4 | Improved |
| add L_fif (ablation) | Pix2LR · BeCoS rendered | acc-hat: 61.7 -> 87.4 | Improved |
| add L_img (ablation) | Pix2LR · BeCoS rendered | acc-hat: 87.4 -> 87.4 | No clear effect |
| use BeCoS instead of FAUST as 3D shape dataset (ablation) | Pix2LR · SPair-71k | acc-hat: 78.2 -> 83.8 | Improved |
| use BeCoS instead of FAUST as 3D shape dataset (ablation) | Pix2LR · BeCoS rendered | acc-hat: 70.7 -> 87.4 | Improved |
| train on both PF-Pascal and SPair-71k images instead of SPair-71k only | Pix2LR · SPair-71k | AVG acc: 82.7 -> 83.8 | Improved |
| train on PF-Pascal only instead of both image datasets | Pix2LR · PF-Pascal | AVG acc: 86.3 -> 86.8 | Improved |
Closest papers by meaning
- Telling Left from Right: Identifying Geometry-Aware Semantic Correspondence (CVPR 2024 · similarity 0.88)
- Left–Right Symmetry Breaking in CLIP-Style Vision-Language Models Trained on Synthetic Spatial-Relation Data (ICML 2026 · similarity 0.87)
- Symmetry Strikes Back: From Single-Image Symmetry Detection to 3D Generation (CVPR 2025 · similarity 0.83)
- SemAlign3D: Semantic Correspondence between RGB-Images through Aligning 3D Object-Class Representations (CVPR 2025 · similarity 0.83)
- Improving Semantic Correspondence with Viewpoint-Guided Spherical Maps (CVPR 2024 · similarity 0.83)
- Large Spatial Model: End-to-end Unposed Images to Semantic 3D (NeurIPS 2024 · similarity 0.83)
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.
