Research map / Detection and segmentation / Tracking
Learning Discriminative Model Prediction for Tracking
Goutam Bhat, Martin Danelljan, Luc Van Gool, Radu Timofte
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 optimizer module using background information (gradient descent) (ablation) | DiMP-18 · OTB-100+NFS+UAV123 combined (ResNet-18) | AUC (%): 58.2 -> 61.6 | Improved |
| use steepest descent instead of gradient descent (ablation) | DiMP-18 · OTB-100+NFS+UAV123 combined (ResNet-18) | AUC (%): 61.6 -> 63.8 | Improved |
| add initializer module (ablation) | DiMP-18 (SD) · OTB-100+NFS+UAV123 combined (ResNet-18) | AUC (%): 58.7 -> 60.0 | Improved |
| train the backbone end to end (ablation) | DiMP-18 · OTB-100+NFS+UAV123 combined (ResNet-18) | AUC (%): 60.0 -> 62.6 | Improved |
| add extra classification conv block (ablation) | DiMP-18 · OTB-100+NFS+UAV123 combined (ResNet-18) | AUC (%): 62.6 -> 63.3 | Improved |
| learn the discriminative loss offline (ablation) | DiMP-18 · OTB-100+NFS+UAV123 combined (ResNet-18) | AUC (%): 63.3 -> 63.8 | Improved |
| update the target model online with a growing training set (ablation) | DiMP-18 · OTB-100+NFS+UAV123 combined (ResNet-18) | AUC (%): 61.7 -> 63.8 | Improved |
| update by model averaging instead of no update (ablation) | DiMP-18 · OTB-100+NFS+UAV123 combined (ResNet-18) | AUC (%): 61.7 -> 61.7 | No clear effect |
| use ResNet-50 instead of ResNet-18 backbone | DiMP · VOT2018 | EAO: 0.402 -> 0.440 | Improved |
| DiMP-50 compared with SiamRPN++ (same backbone) | DiMP-50 · VOT2018 | robustness (failure rate): 0.234 -> 0.153 | Improved |
| use DiMP-18 instead of ATOM | DiMP-18 · GOT10k test (trained on GOT10k only) | AO (%): 55.6 -> 57.9 | Improved |
| use ResNet-50 instead of ResNet-18 | DiMP · GOT10k test (trained on GOT10k only) | AO (%): 57.9 -> 61.1 | 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.
- ATOM — ATOM: Accurate Tracking by Overlap Maximization (CVPR 2019) · better on 6 of 6 shared results
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
- Discriminative and Robust Online Learning for Siamese Visual Tracking (AAAI 2020 · similarity 0.91)
- Target-Aware Deep Tracking (CVPR 2019 · similarity 0.89)
- Learning Feature Embeddings for Discriminant Model based Tracking (ECCV 2020 · similarity 0.88)
- Visual Object Tracking with Discriminative Filters and Siamese Networks: A Survey and Outlook (TPAMI 2022 · similarity 0.88)
- Robust Visual Tracking via Hierarchical Convolutional Features (TPAMI 2018 · similarity 0.88)
- Probabilistic Regression for Visual Tracking (CVPR 2020 · similarity 0.88)
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.
