VS-Sim: A Synthetic Dataset for Viewpoint Shift Robustness

VS-Sim is a synthetic dataset of road scene images that can be used to study the robustness to viewpoint shift of Computer Vision models for several tasks.

This dataset is introduced in the paper: "VS-Sim: A Synthetic Dataset for Viewpoint Shift Robustness", to be presented at ICIAP 2025.

Project page: henriquepm.github.io/VS-SIM/

Dataset Structure

├── Town01/
│   └── town01_multiple_cars/
│       ├── scenario1_14341/
│       │   ├── rgb/
│       │   ├── BEV_semantic_processed/
│       │   ├── PITCHED_Town01/
│       │   ├── PITCHEDDEPTH_Town01/
│       │   ├── LEFT_rgb/
│       │   ├── BOTTOM_rgb/
│       │   ├── ... (other sensor folders)
│       │   ├── gnss.txt
│       │   └── imu.txt
│       └── scenario2_14345/
│           └── ... (similar sensor data)
├── Town02/
│   └── town02_multiple_cars/
│       └── ...
...
└── Town10/
    └── town10_multiple_cars/
        └── ...