Reproducibility package for Frame Blur as an Image-Plane Motion Cue in Counter-UAV and Drone-Tracking Video
Description
This dataset provides the reproducibility package for the manuscript “Frame Blur as an Image-Plane Motion Cue in Counter-UAV and Drone-Tracking Video”. The study tests whether single-frame target blur provides a reliable image-plane motion cue in real counter-UAV and drone-tracking video. Across the studied footage, measured frame-blur length does not show a camera-consistent operational association with motion, while controlled-exposure experiments confirm estimator response only above the observed real-footage range. The package contains Python source code, processed numerical results, diagnostic outputs, environment records, execution commands, file manifests, and SHA-256 checksums supporting the reported analyses, tables, and supplementary results. It covers the primary and alternative blur estimators, bootstrap and robustness analyses, controlled-exposure experiments, temporal-averaging probes, and controlled prediction-sensitivity diagnostics. The original benchmark videos are not redistributed. The analyses use the publicly available Anti-UAV-RGBT, Multi-View Drone Tracking (MVDT), and FRED datasets, which must be obtained from their original sources. Users should begin with reproducibility/REPRODUCIBILITY.md and verify the included SHA-256 manifests before running the scripts.
Files
Steps to reproduce
1. Download the Anti-UAV-RGBT, Multi-View Drone Tracking (MVDT), and FRED benchmark datasets from their original sources cited in the associated manuscript. 2. Download and extract reproducibility_package.zip. Begin with reproducibility/REPRODUCIBILITY.md, which describes the package layout, software environment, dataset-path placeholders, analysis commands, and expected outputs. 3. Use Python 3.10.8 on Ubuntu 22.04.3 or a compatible environment. Install the principal dependencies with: python -m pip install -r requirements.txt For archival reconstruction, use: python -m pip install -r reproducibility/environment/requirements_autodl_20260711_100649.txt 4. Replace the dataset placeholders <Anti-UAV300>, <mvdt>, and <FRED> in the documented commands with local dataset paths. 5. Run the estimator, statistical, exposure, temporal-averaging, and prediction commands listed in reproducibility/REPRODUCIBILITY.md. CPU execution is sufficient for estimator-only analyses; the GRU diagnostics benefit from CUDA. 6. Verify package integrity before use: python reproducibility/verify_sha256.py reproducibility/code_sha256.txt python reproducibility/verify_sha256.py reproducibility/results_sha256.txt The archived environment and authoritative processed outputs are included in the package. The original benchmark videos are not redistributed.
Institutions
- Civil Aviation Flight University of ChinaSichuan, Luocheng