Research map / Architectures and efficiency / CNN design
SiamRPN++: Evolution of Siamese Visual Tracking With Very Deep Networks
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train with spatial-aware random translations to reduce center bias | ResNet-50 Siamese tracker · VOT2018 | EAO: 0.14 | Helped |
| Replace AlexNet with ResNet-50 (ablation) | SiamRPN++ · VOT2018 | EAO: 0.332 -> 0.371 | Helped |
| Fine-tune the backbone (ablation) | ResNet-50 Siamese tracker with L3–L5 and UP-XCorr · VOT2018 | EAO: 0.371 -> 0.390 | Helped |
| Replace UP-XCorr with DW-XCorr (ablation) | ResNet-50 Siamese tracker with L3–L5 and fine-tuning · VOT2018 | EAO: 0.390 -> 0.414 | Helped |
| Aggregate features from L3, L4, and L5 rather than a single layer (ablation) | ResNet-50 SiamRPN++ · VOT2018 | EAO: 4.0% higher than that of the single layer baseline | Helped |
| Use only L5 instead of L4 (ablation) | Fine-tuned ResNet-50 tracker with DW-XCorr · VOT2018 | EAO: 0.374 -> 0.320 | Hurt |
| Aggregate L4 and L5 features (ablation) | Fine-tuned ResNet-50 tracker with DW-XCorr · VOT2018 | EAO: 0.374 -> 0.383 | Helped |
| Aggregate L3 and L4, or L3 and L5, rather than using L4 alone (ablation) | Fine-tuned ResNet-50 tracker with DW-XCorr · VOT2018 | EAO: 0.374 -> 0.346; 0.374 -> 0.336 | Hurt |
| Replace UP-XCorr with DW-XCorr (ablation) | AlexNet Siamese tracker · VOT2018 | EAO: 0.332 -> 0.355 | Helped |
| Increase backbone depth from AlexNet to ResNet-50 (ablation) | SiamRPN++ · OTB2015 | AUC: 0.658 -> 0.696 | Helped |
| Use the deeper SiamRPN++ tracker instead of DaSiamRPN | Visual tracking on OTB2015 · OTB2015 | overlap: 3.8% | Helped |
| Use SiamRPN++ instead of DaSiamRPN | Visual tracking on TrackingNet test · TrackingNet test | AUC: 9.5% | Helped |
Closest papers by meaning
- Deeper and Wider Siamese Networks for Real-Time Visual Tracking (CVPR 2019 · similarity 0.94)
- High Performance Visual Tracking With Siamese Region Proposal Network (CVPR 2018 · similarity 0.92)
- Siamese Box Adaptive Network for Visual Tracking (CVPR 2020 · similarity 0.92)
- SiamCAR: Siamese Fully Convolutional Classification and Regression for Visual Tracking (CVPR 2020 · similarity 0.91)
- Learning To Filter: Siamese Relation Network for Robust Tracking (CVPR 2021 · similarity 0.91)
- HIPTrack: Visual Tracking with Historical Prompts (CVPR 2024 · 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.