Research map / Detection and segmentation / Open-set and few-shot detection
Self-supervised visual learning from interactions with objects
Arthur Aubret, Céline Teulière, Jochen Triesch
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 action-aware loss aligning action embedding with the pair of view representations | SimCLR · RT4K | test accuracy (%): 71.1 -> 74.8 | Improved |
| add action-aware loss | SimCLR · MVImgNet-T (2% of objects) | test accuracy (%): 40.1 -> 42.8 | Improved |
| add action-aware loss on the huge MVImgNet-F dataset | SimCLR · MVImgNet-F | test accuracy (%): 96.4 (TT) vs 96.1 (AA) | No clear effect |
| add action-aware loss to BYOL | BYOL · RT4K | test accuracy (%): 69% -> 71.3% | Improved |
| pretrain on MVImgNet-F with AA-SimCLR and transfer | SimCLR · ImageNet-1K linear probe | accuracy (%): 40.3 -> 44.3 | Improved |
| AA-SimCLR vs SimCLR-CIPER on Tiny-ImageNet transfer | SimCLR · Tiny-ImageNet linear probe | accuracy (%): 38.6 (CIPER) vs 38 (AA) | No clear effect |
| use smaller (1x256) projection heads (ablation) | AA-SimCLR · MVImgNet-S | test accuracy (%): AA: 78.3 -> 73.4; TT: 77.3 -> 68.6 | Mixed |
| remove color augmentations during training (ablation) | SimCLR-TT vs AA-SimCLR · RT4K | category accuracy | Improved |
| predict the learned action representation rather than the raw action (ablation) | AA-SSL vs CIPER · RT4K chair PaCMAP | view-wise alignment of objects | Improved |
Closest papers by meaning
- Learning State-Aware Visual Representations from Audible Interactions (NeurIPS 2022 · similarity 0.89)
- Time to augment self-supervised visual representation learning (ICLR 2023 · similarity 0.89)
- Revisiting Self-Supervised Visual Representation Learning (CVPR 2019 · similarity 0.89)
- Object Concepts Emerge from Motion (NeurIPS 2025 · similarity 0.88)
- No More Shortcuts: Realizing the Potential of Temporal Self-Supervision (AAAI 2024 · similarity 0.87)
- Probing the Mid-level Vision Capabilities of Self-Supervised Learning (CVPR 2025 · similarity 0.87)
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.
