Research map / Detection and segmentation / Tracking
Bridging Images and Videos: A Simple Learning Framework for Large Vocabulary Video Object Detection
Sanghyun Woo, Kwanyong Park, Seoung Wug Oh, In So Kweon, Joon-Young Lee
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 SimLearn unified learning | FasterRCNN-RFS* · TAO-val | Track AP50: 13.4 -> 19.6 | Improved |
| add SimLearn unified learning | CenterNet2-RFS · TAO-val | Track AP50: 18.9 -> 21.6 | Improved |
| add SimLearn unified learning | FasterRCNN-Seesaw · TAO-val | Track AP50: 15.4 -> 20.2 | Improved |
| use strong zoom-in/out scale jittering (0.1-2.0) instead of small range (ablation) | LVIS-tracker · TAO-val | Track AP50: 13.2 -> 14.4 | Improved |
| add mosaicing to zoom in/out (ablation) | LVIS-tracker · TAO-val | Track AP50: 14.4 -> 17.8 | Improved |
| add mosaicing augmentation alone (ablation) | LVIS-tracker · TAO-val | Box AP: 19.0 -> 16.2 | Hurt |
| add pseudo-label distillation and negative correction to vanilla teacher-student (ablation) | Teacher-student · TAO-val | Track AP50: 15.9 -> 18.5 | Improved |
| use soft pseudo labels instead of hard (ablation) | Teacher-student · TAO-val | Track AP50: 15.9 -> 18.5 | Improved |
| use MSE instead of KL distillation loss (ablation) | Teacher-student · TAO-val | Track AP50: 16.8 -> 18.5 | Improved |
| add semantic consistency loss (ablation) | Teacher-student tracker · TAO-val | Track AP50: 18.5 -> 19.5 | Improved |
| avoid naive fine-tuning on TAO by using teacher-student (ablation) | LVIS-tracker · TAO-val | Track AP50: 11.4 -> 18.5 | Improved |
| use teacher-student instead of track-only fine-tuning | LVIS to TAO transfer · TAO-val NEW classes | Track AP50: 16.9 -> 20.6 | Improved |
Closest papers by meaning
- TAO: A Large-Scale Benchmark for Tracking Any Object (ECCV 2020 · similarity 0.91)
- "A Simple Approach and Benchmark for 21,000-Category Object Detection" (ECCV 2022 · similarity 0.90)
- OVT-B: A New Large-Scale Benchmark for Open-Vocabulary Multi-Object Tracking (NeurIPS 2024 · similarity 0.90)
- TrackingNet: A Large-Scale Dataset and Benchmark for Object Tracking in the Wild (ECCV 2018 · similarity 0.89)
- COVTrack++: Learning Open-Vocabulary Multi-Object Tracking from Continuous Videos via a Synergistic Paradigm (TPAMI 2026 · similarity 0.89)
- VastTrack: Vast Category Visual Object Tracking (NeurIPS 2024 · 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.
