Featured research / Dataset
York-Tank
York-Tank: A Degradation-Aware and Difficulty-Stratified Stereo Dataset for Underwater Visual SLAM
York-Tank is a controlled underwater stereo dataset for investigating how visual degradation affects SLAM tracking, localization, and failure behavior. Sequences share a controlled tank scene and are organized by dominant degradation factor and difficulty, supporting structured comparisons across challenging underwater conditions.
Complete dataset publicly available. Zenodo v2 includes all 16 sequences and 18,948 stereo image pairs, distributed as 16 sequence archives and one support archive (approximately 51.1 GB). Download the complete release. Availability verified on .
- Sequences
- 16
- Stereo image pairs
- 18,948
- Degradation factors
- 5
- Difficulty levels
- 4
Rectified grayscale stereo · 1280 × 800 pixels · 20 Hz · Approximately 15.8 minutes
Dataset characteristics
- Controlled scene: a shared tank environment for studying the effects of visual degradation.
- Five factors: low illumination, non-uniform illumination, fast motion, texture deficiency, and dynamic interference.
- Structured difficulty: baseline, easy, hard, and hybrid conditions.
- Release contents: synchronized stereo images, IMU measurements, calibration files, raw AprilTag-array reference poses, and ROS bag files.
Difficulty structure
| Level | Condition | Sequences |
|---|---|---|
| Level 1 | Baseline | 2 |
| Level 2 | Easy single degradation | 5 |
| Level 3 | Hard single degradation | 5 |
| Level 4 | Hybrid degradation | 4 |