Research map / Detection and segmentation / Tracking
TrackingNet: A Large-Scale Dataset and Benchmark for Object Tracking in the Wild
Matthias Muller, Adel Bibi, Silvio Giancola, Salman Alsubaihi, Bernard Ghanem
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 linear-decay weighted forward/backward average instead of forward pass only (ablation) | ECO · OTB100 1-sec sequences | OPE success: 0.805 -> 0.843 | Improved |
| use linear-decay weighted average instead of forward pass only (ablation) | DCF · OTB100 1-sec sequences | OPE success: 0.771 -> 0.825 | Improved |
| use linear-decay weighted average instead of forward pass only (ablation) | STAPLE CA · OTB100 1-sec sequences | OPE success: 0.799 -> 0.841 | Improved |
| use linear weighting instead of plain 0.5 average of passes (ablation) | ECO · OTB100 1-sec sequences | OPE success: 0.824 -> 0.843 | Improved |
| use linear weighting instead of exponential decay (ablation) | STAPLE CA · OTB100 1-sec sequences | OPE success: 0.835 -> 0.841 | No clear effect |
| combine forward and backward pass (average) instead of forward only (ablation) | DCF · OTB100 1-sec sequences | OPE success: 0.771 -> 0.799 | Improved |
| fine-tune on one of twelve TrackingNet chunks (2511 videos) | SiameseFC · OTB100 | success: 0.569 -> 0.576 | Improved |
| fine-tune on one TrackingNet chunk | SiameseFC · OTB100 | precision: 0.765 -> 0.781 | Improved |
| fine-tune on one TrackingNet chunk | SiameseFC · TrackingNet Test | success: 0.571 -> 0.581 | Improved |
| fine-tune on one TrackingNet chunk | SiameseFC · TrackingNet Test | precision: 0.533 -> 0.543 | Improved |
| fine-tune on one TrackingNet chunk | SiameseFC · TrackingNet Test | normalized precision: 0.663 -> 0.673 | Improved |
| subsample frames by tracker speed (real-time constraint) | Struck · TrackingNet Test | success: as much as 50% | Hurt |
Closest papers by meaning
- TAO: A Large-Scale Benchmark for Tracking Any Object (ECCV 2020 · similarity 0.95)
- GOT-10k: A Large High-Diversity Benchmark for Generic Object Tracking in the Wild (TPAMI 2019 · similarity 0.93)
- VastTrack: Vast Category Visual Object Tracking (NeurIPS 2024 · similarity 0.93)
- LaSOT: A High-quality Large-scale Single Object Tracking Benchmark (IJCV 2020 · similarity 0.93)
- Long-term Tracking in the Wild: a Benchmark (ECCV 2018 · similarity 0.92)
- OVT-B: A New Large-Scale Benchmark for Open-Vocabulary Multi-Object Tracking (NeurIPS 2024 · similarity 0.90)
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.
