54,120 frames of multi-sensor data including camera, radar, LiDAR, and GPS/IMU for inland water surface perception. The dataset provides annotations for object detection, semantic segmentation, and instance segmentation across 7 categories such as boats, humans, and buoys.
Use Cases
- Train multi-modal object detection models using the radar point cloud and RGB image channels
- Develop semantic segmentation algorithms to classify 'water' and 'land' pixels for autonomous navigation
- Benchmark sensor fusion techniques for obstacle avoidance using the synchronized LiDAR and radar data
Strengths
- 54,120 temporal frames with synchronized camera, radar, LiDAR, and GPS/IMU data
- 2D and 3D bounding box annotations for 7 object classes including 'boat', 'pier', and 'buoy'
- Pixel-level semantic masks for 'water', 'land', and 'object' segmentation
- Data captured across diverse environmental conditions including night, rain, and fog