Research map / 3D vision / Depth and stereo
ARKitTrack: A New Diverse Dataset for Tracking Using Mobile RGB-D Data
Haojie Zhao, Junsong Chen, Lijun Wang, Huchuan Lu
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 |
|---|---|---|---|
| Fine-tune the RGB tracker on ARKitTrack training data using RGB channels (ablation) | OSTrack · ARKitTrack-VOT | F-score: 0.441 -> 0.443 | Improved |
| Fine-tune the RGB tracker on DepthTrack training data using RGB channels (ablation) | OSTrack · DepthTrack | F-score: 0.567 -> 0.574 | Improved |
| Use BEV features without cross-view fusion (ablation) | Fine-tuned OSTrack · ARKitTrack-VOT | F-score: 0.443 -> 0.423 | Hurt |
| Use BEV features without cross-view fusion (ablation) | Fine-tuned OSTrack · DepthTrack | F-score: 0.574 -> 0.553 | Hurt |
| Add BEV geometry and cross-view fusion (ablation) | Fine-tuned OSTrack · ARKitTrack-VOT | F-score: 0.443 -> 0.470 | Improved |
| Add BEV geometry and cross-view fusion (ablation) | Fine-tuned OSTrack · DepthTrack | F-score: 0.574 -> 0.602 | Improved |
| Model depth for each pixel with a Gaussian distribution centered on its input depth (ablation) | BEV transformation with cross-view fusion · ARKitTrack-VOT | F-score: 0.470 -> 0.479 | Improved |
| Model depth for each pixel with a Gaussian distribution centered on its input depth (ablation) | BEV transformation with cross-view fusion · DepthTrack | F-score: 0.602 -> 0.612 | Improved |
| Use one dynamic template updated every N frames (ablation) | VOS baseline without dynamic templates · ARKitTrack-VOS | J&F: 0.630 -> 0.642 | Improved |
| Use IoU prediction to control updates to one template (ablation) | VOS model with ONE · ARKitTrack-VOS | J&F: 0.642 -> 0.648 | Improved |
| Use IoU prediction to control adding templates to the memory bank (ablation) | VOS model · ARKitTrack-VOS | J&F: 0.648 -> 0.662 | Improved |
| Train and evaluate on the diverse ARKitTrack benchmark rather than existing RGB-D benchmarks | RGB-D trackers · ARKitTrack-VOT | F-score | Mixed |
Closest papers by meaning
- Single-Model and Any-Modality for Video Object Tracking (CVPR 2024 · similarity 0.86)
- SpatialTracker: Tracking Any 2D Pixels in 3D Space (CVPR 2024 · similarity 0.85)
- CSTrack: Enhancing RGB-X Tracking via Compact Spatiotemporal Features (ICML 2025 · similarity 0.84)
- Temporal Adaptive RGBT Tracking with Modality Prompt (AAAI 2024 · similarity 0.84)
- TAPVid-3D: A Benchmark for Tracking Any Point in 3D (NeurIPS 2024 · similarity 0.84)
- Collaborating Vision, Depth, and Thermal Signals for Multi-Modal Tracking: Dataset and Algorithm (NeurIPS 2025 · similarity 0.84)
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.
