Research map / Video / Video segmentation and tracking
Tracking Emerges by Colorizing Videos
Carl Vondrick, Abhinav Shrivastava, Alireza Fathi, Sergio Guadarrama, Kevin Murphy
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 |
|---|---|---|---|
| use video colorization embeddings instead of single-image colorization embeddings (ablation) | pointer-based tracker · DAVIS 2017 val | segment overlap (mean J): 4.7 -> 34.6 | Improved |
| use learned colorization pointer instead of FlowNet2 optical flow warping | self-supervised tracker · DAVIS 2017 val | segment overlap (mean J): 26.7 -> 34.6 | Improved |
| use learned colorization pointer instead of FlowNet2 optical flow warping | self-supervised tracker · DAVIS 2017 val | boundary accuracy (mean F): 25.2 -> 32.7 | Improved |
| use learned pointer instead of optical flow on multi-object videos | self-supervised tracker · DAVIS 2017 val (2-5 objects) | segment overlap (J): 19.1 -> 31.0 | Improved |
| use colorization tracker instead of FlowNet2 at strict threshold | self-supervised tracker · JHMDB | PCK@.1: 45.2 -> 45.2 | No clear effect |
| use colorization tracker instead of FlowNet2 at relaxed threshold | self-supervised tracker · JHMDB | PCK@.5: 85.5 -> 91.4 | Improved |
| use colorization tracker instead of identity baseline at strict threshold | self-supervised tracker · JHMDB | PCK@.1: 43.1 -> 45.2 | Improved |
| add full supervision (ImageNet, COCO, DAVIS) instead of self-supervision | tracker · DAVIS 2017 val | segment overlap (mean J): 34.6 -> 55.1 | Improved |
| lower softmax temperature at inference to T=0.5 | colorization tracker · DAVIS 2017 val | label sharpness | No clear effect |
Closest papers by meaning
- Video Diffusion Models Excel at Tracking Similar-Looking Objects Without Supervision (NeurIPS 2025 · similarity 0.89)
- Colorization as a Proxy Task for Visual Understanding (CVPR 2017 · similarity 0.87)
- Track, Check, Repeat: An EM Approach to Unsupervised Tracking (CVPR 2021 · similarity 0.87)
- Track4Gen: Teaching Video Diffusion Models to Track Points Improves Video Generation (CVPR 2025 · similarity 0.85)
- Single-Model and Any-Modality for Video Object Tracking (CVPR 2024 · similarity 0.85)
- OneTracker: Unifying Visual Object Tracking with Foundation Models and Efficient Tuning (CVPR 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.
