Research map / Detection and segmentation / Tracking
Learning to Track at 100 FPS with Deep Regression Networks
David Held, Sebastian Thrun, Silvio Savarese
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 L1 loss instead of L2 loss (ablation) | GOTURN · VOT 2014 | overall error: 0.43 -> 0.24 | Improved |
| add Laplace-distributed random crops (motion smoothness) instead of uniform crops (ablation) | GOTURN · VOT 2014 | overall error: 0.30 -> 0.24 | Improved |
| add video training data to image-only training (ablation) | GOTURN · VOT 2014 | overall error: 0.35 -> 0.24 | Improved |
| add still-image training data to video-only training (ablation) | GOTURN · VOT 2014 | overall error: 0.29 -> 0.24 | Improved |
| feed the previous frame as additional input (ablation) | GOTURN · VOT 2014 (occlusion, camera motion) | overall error | Improved |
| feed the previous frame as additional input (ablation) | GOTURN · VOT 2014 (size change, no variation) | overall error | Hurt |
| increase number of training videos | GOTURN · VOT 2014 | overall rank | Improved |
| add online fine-tuning at test time | GOTURN · VOT 2014 | overall rank | No clear effect |
| use 3 fully-connected layers instead of 2 or 4 (ablation) | GOTURN · VOT 2014 | tracking error | No clear effect |
| increase augmented images per batch of 50 (up to 49) (ablation) | GOTURN · VOT 2014 | tracking error | Improved |
| add training videos of the same class as test objects (ablation) | GOTURN · VOT 2014 (classes with at least 25 training videos) | overall error | Improved |
Closest papers by meaning
- TrackingNet: A Large-Scale Dataset and Benchmark for Object Tracking in the Wild (ECCV 2018 · similarity 0.88)
- Robust Object Tracking Based on Temporal and Spatial Deep Networks (ICCV 2017 · similarity 0.87)
- Learning Policies for Adaptive Tracking With Deep Feature Cascades (ICCV 2017 · similarity 0.87)
- Need for Speed: A Benchmark for Higher Frame Rate Object Tracking (ICCV 2017 · similarity 0.87)
- Learning Regression and Verification Networks for Robust Long-term Tracking (IJCV 2021 · similarity 0.86)
- Learning Multi-Domain Convolutional Neural Networks for Visual Tracking (CVPR 2016 · similarity 0.86)
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.
