Research map / Architectures and efficiency / Attention and transformers
HiFT: Hierarchical Feature Transformer for Aerial Tracking
Ziang Cao, Changhong Fu, Junjie Ye, Bowen Li, Yiming Li
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 original transformer with object queries (ablation) | Baseline · UAV20L | precision: 0.611 -> 0.597 | Hurt |
| add original transformer with object queries (ablation) | Baseline · UAV20L | success: 0.463 -> 0.446 | Hurt |
| replace object queries with feature map in transformer (FT) (ablation) | Baseline · UAV20L | precision: 0.611 -> 0.675 | Improved |
| replace object queries with feature map in transformer (FT) (ablation) | Baseline · UAV20L | success: 0.463 -> 0.496 | Improved |
| add modulation layer to feature transformer (FT to HFT) (ablation) | Baseline+FT · UAV20L | precision: 0.675 -> 0.763 | Improved |
| add modulation layer to feature transformer (FT to HFT) (ablation) | Baseline+FT · UAV20L | success: 0.496 -> 0.566 | Improved |
| omit positional encoding on M5 (vs direct PE) (ablation) | HiFT · UAV20L | precision: 0.689 -> 0.763 | Improved |
| omit positional encoding on M5 (vs direct PE) (ablation) | HiFT · UAV20L | success: 0.523 -> 0.566 | Improved |
| use circular classification label instead of rectangle label (ablation) | HiFT · UAV20L | precision: 0.629 -> 0.763 | Improved |
| use circular classification label instead of rectangle label (ablation) | HiFT · UAV20L | success: 0.486 -> 0.566 | Improved |
| HiFT (AlexNet) vs SiamRPN++ (ResNet-50) deeper backbone | HiFT · Four aerial benchmarks average | avg precision: 0.774 -> 0.783 | Improved |
| HiFT vs SiamRPN++ on UAV20L precision | SiamRPN++ · UAV20L | precision: 0.696 -> 0.763 | 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.
- SiamRPN++ — SiamRPN++: Evolution of Siamese Visual Tracking With Very Deep Networks (CVPR 2019) · better on 8 of 8 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
- Hiera: A Hierarchical Vision Transformer without the Bells-and-Whistles (ICML 2023 · similarity 0.86)
- Robust Visual Tracking via Hierarchical Convolutional Features (TPAMI 2018 · similarity 0.85)
- Similarity-Guided Layer-Adaptive Vision Transformer for UAV Tracking (CVPR 2025 · similarity 0.85)
- From Coarse to Fine: Robust Hierarchical Localization at Large Scale (CVPR 2019 · similarity 0.85)
- Learning Adaptive and View-Invariant Vision Transformer for Real-Time UAV Tracking (ICML 2024 · similarity 0.84)
- Target Focused Shallow Transformer Framework for Efficient Visual Tracking (AAAI 2024 · similarity 0.83)
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.
