Research map / Detection and segmentation / Tracking
Discriminative Correlation Filter Tracker with Channel and Spatial Reliability
Alan Lukežič, Tomáš Vojíř, Luka Čehovin Zajc, Jiří Matas, Matej Kristan
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 channel reliability weights (ablation) | CSR-DCF · VOT2016 | EAO: 0.297 -> 0.338 | Improved |
| use segmentation-based spatial reliability map instead of uniform box map (ablation) | CSR-DCF · VOT2016 | EAO: 0.264 -> 0.338 | Improved |
| add both channel and spatial reliability (vs uniform map and no channel weights) (ablation) | CSR-DCF · VOT2016 | EAO: 0.256 -> 0.338 | Improved |
| add a uniform spatial reliability map to a plain DCF (no channel weights) (ablation) | CSR-DCF without reliability · VOT2016 | EAO: 0.152 -> 0.256 | Improved |
| use ADMM constrained optimization instead of closed-form solution followed by masking (ablation) | CSR-DCF without channel weights · VOT2016 | EAO: 0.251 -> 0.297 | Improved |
| use proposed spatial reliability constraint instead of spatial regularization or limited boundaries | single-scale HoG correlation filter · VOT2015 | AUC: 0.32 vs 0.28 (T_SR) and 0.16 (T_LB) | Improved |
| replace estimated reliability map with ground-truth segmentation masks (ablation) | CSR-DCF · VOT2016 | EAO: 0.338 -> 0.338 | No clear effect |
| use CSR-DCF constraints instead of SRDCF/LBCF for non-axis-aligned initialization | HoG correlation filter · VOT2015 non-axis-aligned initialization | Gamma_frm: 221 vs 95 (SRDCF) and 37 (LBCF) | Improved |
| replace deep features with simple HoG/Colornames features in CCOT | CCOT · VOT2016 | EAO: 0.331 -> 0.274 | Hurt |
| add channel and spatial reliability to simple-feature DCF versus SRDCF | CSR-DCF · VOT2016 | EAO: 0.247 -> 0.338 | 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.
- Staple — Staple: Complementary Learners for Real-Time Tracking (CVPR 2016) · better on 1 of 1 shared result
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 Correlation Filter With Channel and Spatial Reliability (CVPR 2017 · similarity 0.99)
- Correlation Tracking via Joint Discrimination and Reliability Learning (CVPR 2018 · similarity 0.91)
- Beyond Correlation Filters: Learning Continuous Convolution Operators for Visual Tracking (ECCV 2016 · similarity 0.90)
- ECO: Efficient Convolution Operators for Tracking (CVPR 2017 · similarity 0.90)
- Joint Group Feature Selection and Discriminative Filter Learning for Robust Visual Object Tracking (ICCV 2019 · similarity 0.89)
- Adaptive Channel Selection for Robust Visual Object Tracking with Discriminative Correlation Filters (IJCV 2021 · similarity 0.89)
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.
