Research map / Detection and segmentation / Open-set and few-shot detection
Self-Supervised Learning via Conditional Motion Propagation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use conditional motion propagation pretraining instead of the strongest listed self-supervised baseline | AlexNet representation for semantic segmentation · PASCAL VOC 2012 semantic segmentation | mIoU: 41.4 -> 44.5 | Helped |
| Use conditional motion propagation pretraining instead of the strongest listed self-supervised baseline | ResNet-50 representation for semantic segmentation · PASCAL VOC 2012 semantic segmentation | mIoU: 54.6 -> 59.0 | Helped |
| Use conditional motion propagation pretraining instead of the strongest listed self-supervised baseline | ResNet-50 representation for instance detection · COCO 2017 instance segmentation | bounding box mAP: 31.5 -> 32.3 | Helped |
| Use conditional motion propagation pretraining instead of the strongest listed self-supervised baseline | ResNet-50 representation for instance segmentation · COCO 2017 instance segmentation | mask mAP: 29.2 -> 29.8 | Helped |
| Use conditional motion propagation pretraining instead of the strongest listed self-supervised baseline | ResNet-50 representation for single-person parsing · LIP Single-Person Parsing | mIoU: 36.7 -> 40.2 | Helped |
| Use conditional motion propagation pretraining instead of the strongest listed self-supervised baseline | ResNet-50 representation for multi-person parsing · LIP Multi-Person Parsing | mIoU: 52.5 -> 52.9 | Helped |
| Combine CMP and ImageNet-pretrained models | LIP human parsing model · LIP Single-Person Parsing | mIoU: 40.2 -> 42.9 | Helped |
| Combine CMP and ImageNet-pretrained models | LIP human parsing model · LIP Multi-Person Parsing | mIoU: 52.9 -> 55.8 | Helped |
| Use CMP rather than Wang et al.'s self-supervised pretraining | VGG-16 representation · PASCAL VOC 2007 detection | mAP: 63.2 -> 56.8 | Hurt |
| Use CMP rather than Wang et al.'s self-supervised pretraining | VGG-16 representation · PASCAL VOC 2012 segmentation | mIoU: 54.0 -> 57.6 | Helped |
| Use CMP for interactive annotation rather than Polygon RNN++ | Semi-automatic pixel-level annotation · COCO subset | time per instance: 17.6s -> 10.2s | Helped |
| Use CMP for interactive annotation rather than Polygon RNN++ | Semi-automatic pixel-level annotation · COCO subset | failed instances: 25/170 -> 0/170 | Helped |
Closest papers by meaning
- ProMotion: Prototypes As Motion Learners (CVPR 2024 · similarity 0.86)
- H-Flow: Self-supervised Human Scene Flow via Physics-inspired Joint Multi-modal Learning (NeurIPS 2026 · similarity 0.86)
- Visual Representation Learning with Stochastic Frame Prediction (ICML 2024 · similarity 0.85)
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 2025 · similarity 0.85)
- Self-Supervised Learning of Motion Concepts by Optimizing Counterfactuals (NeurIPS 2025 · similarity 0.85)
- Learning Segmentation from Point Trajectories (NeurIPS 2024 · similarity 0.85)
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.