XS-VID

Dataset Overview

Dataset Name

XS‑VID

Dataset Download

• Google Drive:
  • Annotations: Link
  • Images (0‑3): Link
  • Images (4‑5): Link
• BaiduNetDisk: Link

Dataset Format

• Annotation Formats: COCO, COCOVID, YOLO

Dataset Organization

data_root_dir/
├── test.txt
├── train.txt
├── images/
│   ├── video1/
│   │   ├── 0000000.png
│   │   └── 0000001.png
│   ├── video2/
│   │   └── ...
│   └── ...
└── labels/
    ├── video1/
    │   ├── 0000000.txt
    │   └── 0000001.txt
    ├── video2/
    │   └── ...
    └── ...

Results

• XS‑VID: Provides performance metrics for multiple methods, including $AP$, $AP_{50}$, $AP_{75}$, $AP_{eS}$, $AP_{rS}$, $AP_{gS}$, and Inference (ms).
• Visdrone2019 VID (test‑dev): Provides performance metrics for multiple methods, including $AP$, $AP_{50}$, $AP_{75}$, $AP_{eS}$, $AP_{rS}$, $AP_{gS}$, $AP_{m}$, and $AP_{l}$.

Checkpoints

• YOLOFT‑L: Parameters 45.16 M, FLOPs 230.14 G, inference time 36 ms, dataset XS‑VID, checkpoint Link
• YOLOFT‑S: Parameters 53.58 M, FLOPs 13.02 G, inference time 16 ms, dataset XS‑VID, checkpoint Link