How tjubiit/tju-dhd works
This repository presents the TJU-DHD research dataset, not a software runtime. The documented user workflow is to choose a scene and task, obtain the corresponding images and annotations from the listed hosts, then evaluate detectors using the benchmark protocols and tools. The README also describes high-resolution, diverse data and pedestrian same-scene and cross-scene evaluation. No source excerpts are available, so links reflect documented capabilities only.
Main entry points: README.md Dataset guide
Components 13
Dataset collections
- Dataset guide
README.md - Object detection data
README.md - Pedestrian data
README.md - Traffic scene
README.md - Campus scene
README.md - Images
README.md - Annotations
README.md
Evaluation benchmarks
- Detection benchmarks
README.md - Pedestrian evaluation
README.md - COCO evaluation tools
Dataset distribution
- OneDrive
- Baidu Netdisk
Other components
- Dataset user
Connections 15
- Dataset user → Dataset guide: consults
README.md - Dataset guide → Object detection data: documents
README.md - Dataset guide → Pedestrian data: documents
README.md - Object detection data → Traffic scene: covers
README.md - Object detection data → Campus scene: covers
README.md - Pedestrian data → Traffic scene: covers
README.md - Pedestrian data → Campus scene: covers
README.md - Object detection data → Images: includes
README.md - Object detection data → Annotations: includes
README.md - Dataset guide → Detection benchmarks: describes
README.md - Dataset guide → Pedestrian evaluation: describes
README.md - Detection benchmarks → COCO evaluation tools: uses
README.md - Dataset guide → OneDrive: links downloads
README.md - Dataset guide → Baidu Netdisk: links downloads
README.md - Pedestrian evaluation → Detection benchmarks: supports cross-scene
README.md