Research map / Detection and segmentation / Tracking
Joint Representation and Truncated Inference Learning for Correlation Filter based Tracking
Yingjie Yao, Xiaohe Wu, Lei Zhang, Shiguang Shan, Wangmeng Zuo
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 truncated inference learning (updater network) to fixed VGG-M features (ablation) | BACF-VGGM · OTB-2015 | AUC: 63.1 -> 66.5 | Improved |
| add joint feature representation learning to RTINet-VGGM (ablation) | RTINet-VGGM · OTB-2015 | AUC: 66.5 -> 68.2 | Improved |
| add learned representation to BACF with VGG-M (ablation) | BACF-VGGM · OTB-2015 | AUC: 63.1 -> 64.0 | Improved |
| replace HOG features with VGG-M features in BACF (ablation) | BACF · OTB-2015 | AUC: 61.5 -> 63.1 | Improved |
| use RTINet-learned features in CFNet (ablation) | CFNet · OTB-2015 | AUC: 56.8 -> 58.0 | Improved |
| use RTINet-learned features in CFNet-conv1 (ablation) | CFNet-conv1 · OTB-2015 | AUC: 53.6 -> 54.8 | Improved |
| integrate BACF unrolled ADMM inference and joint learning in place of CFNet (ablation) | CFNet · OTB-2015 | AUC: 56.8 -> 68.2 | Improved |
| use three conv layers instead of conv1 only in RTINet (ablation) | RTINet-conv1 · OTB-2015 | AUC: 64.3 -> 68.2 | Improved |
| increase stages from one to two (ablation) | RTINet · OTB-2015 | AUC: 59.6 -> 68.2 | Improved |
| increase stages beyond two (up to ten) (ablation) | RTINet · OTB-2015 | AUC: 68.2 -> 66.2 | Hurt |
| train RTINet with a fixed stage count of 10 versus train small K directly (ablation) | RTINet · OTB-2015 | AUC | Hurt |
| RTINet versus ECO on VOT2016 accuracy and robustness | RTINet · VOT2016 | accuracy / robustness: 0.57 vs 0.54 accuracy; 1.07 vs 0.72 robustness | Mixed |
Closest papers by meaning
- End-To-End Representation Learning for Correlation Filter Based Tracking (CVPR 2017 · similarity 0.89)
- Learning Background-Aware Correlation Filters for Visual Tracking (ICCV 2017 · similarity 0.88)
- Correlation Tracking via Joint Discrimination and Reliability Learning (CVPR 2018 · similarity 0.88)
- ECO: Efficient Convolution Operators for Tracking (CVPR 2017 · similarity 0.87)
- CREST: Convolutional Residual Learning for Visual Tracking (ICCV 2017 · similarity 0.87)
- Context-Aware Correlation Filter Tracking (CVPR 2017 · 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.
