Research map / Detection and segmentation / Open-set and few-shot detection
Self-Supervised Viewpoint Learning From Image Collections
Siva Karthik Mustikovela, Varun Jampani, Shalini De Mello, Sifei Liu, Umar Iqbal, Carsten Rother, Jan Kautz
Check 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 image consistency loss (ablation) | SSV viewpoint network · BIWI | MAE (degrees): 7.2 -> 6.7 | Improved |
| Add symmetry loss alongside image consistency (ablation) | SSV viewpoint network · BIWI | MAE (degrees): 8.3 -> 6.7 | Improved |
| Add paired style and viewpoint loss when training the synthesis network (ablation) | SSV · BIWI | MAE (degrees): 7.7 -> 6.7 | Improved |
| Refine using unlabeled images from the target BIWI domain | SSV-Full · BIWI held-out sequences | MAE (degrees): 6.8 -> 5.8 | Improved |
| Use SSV instead of HoloGAN with viewpoint output | Pascal3D+ car viewpoint estimation · Car | Median geodesic error (degrees): 16.3 -> 10.1 | Improved |
| Use SSV instead of HoloGAN with viewpoint output | Pascal3D+ bus viewpoint estimation · Bus | Median geodesic error (degrees): 14.2 -> 9.0 | Improved |
| Use SSV instead of HoloGAN with viewpoint output | Pascal3D+ train viewpoint estimation · Train | Median geodesic error (degrees): 9.7 -> 5.3 | Improved |
| Train without viewpoint annotations, using image collections and synthesis-based self-supervision | SSV-Full · Pascal3D+ test, Train | Median geodesic error (degrees): 6.1 -> 5.3 | Improved |
Closest papers by meaning
- ViewNet: Unsupervised Viewpoint Estimation From Conditional Generation (ICCV 2021 · similarity 0.89)
- Pose-Aware Self-Supervised Learning with Viewpoint Trajectory Regularization (ECCV 2024 · similarity 0.88)
- Demystifying Contrastive Self-Supervised Learning: Invariances, Augmentations and Dataset Biases (NeurIPS 2020 · similarity 0.88)
- ShapeCodes: Self-Supervised Feature Learning by Lifting Views to Viewgrids (ECCV 2018 · similarity 0.87)
- Self-Supervised Learning of Depth Inference for Multi-View Stereo (CVPR 2021 · similarity 0.86)
- Beyond Random Augmentations: Pretraining with Hard Views (ICLR 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.
