Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-supervised Pre-training for Mirror Detection
Jiaying Lin, Rynson W.H. Lau
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 three-stage progressive SSL pre-training (image, patch, pixel) versus supervised ImageNet pre-training | MirrorNet · MSD test | F-measure: 0.727 -> 0.763 | Improved |
| add image-level pre-training only versus random init (ablation) | MirrorNet · MSD test | F-measure: 0.601 -> 0.623 | No clear effect |
| add patch-level pre-training versus random init (ablation) | MirrorNet · MSD test | F-measure: 0.601 -> 0.689 | Improved |
| add image-level stage to patch-level pre-training (ablation) | MirrorNet · MSD test | F-measure: 0.689 -> 0.723 | Improved |
| add pixel-level stage to patch-level pre-training (ablation) | MirrorNet · MSD test | MAE: 0.162 -> 0.124 | Improved |
| add image-level stage to patch + pixel pre-training (ablation) | MirrorNet · MSD test | F-measure: 0.758 -> 0.763 | Improved |
| add location predictor in patch-level stage (ablation) | MirrorNet · MSD test | F-measure: 0.731 -> 0.763 | Improved |
| train the three stages sequentially instead of in parallel (ablation) | MirrorNet · MSD test | MAE: 0.331 -> 0.116 | Improved |
| increase pre-training iterations beyond 20,000 per stage (ablation) | SSL pre-training framework · MSD test | F-measure | Hurt |
| use our SSL pre-training instead of generic SSL methods on VCNet | VCNet · MSD test | F-measure: 0.857 -> 0.886 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Barlow Twins — Barlow Twins: Self-Supervised Learning via Redundancy Reduction (ICML 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Detect Any Mirrors: Boosting Learning Reliability on Large-Scale Unlabeled Data with an Iterative Data Engine (CVPR 2025 · similarity 0.90)
- ZOOM: Learning Video Mirror Detection with Extremely-Weak Supervision (AAAI 2024 · similarity 0.90)
- Weakly-Supervised Mirror Detection via Scribble Annotations (AAAI 2024 · similarity 0.89)
- Effective Video Mirror Detection with Inconsistent Motion Cues (CVPR 2024 · similarity 0.84)
- Probing the Mid-level Vision Capabilities of Self-Supervised Learning (CVPR 2025 · similarity 0.84)
- Beyond Random Augmentations: Pretraining with Hard Views (ICLR 2025 · 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.
