AI in maritime - Autonomous ships
summary
Camera footage from an LNG tanker and 14,000 annotated images (plus synthetic augmentations) train sea‑state classifiers and object detectors; a visual transformer (A3) reaches about 64% accuracy across Beaufort classes 1–8 when real and synthetic data are combined, while YOLOv8 fine‑tuned with 100–1,000 annotated images improves small‑object detection (boats, buoys, swimmers). Models run on GPUs, show real‑time demos, and aim to fuse radar/LiDAR and thermal sensors for tougher lighting and weather.
AI-written summary · about
chapters (13)
- 0:00AI in maritime
- 3:21Smart shipping trends
- 5:56AI methods for maritime ops
- 7:00Autonomous ship examples
- 8:24Partially autonomous systems
- 9:48InnoMaritime project
- 10:51Sea state estimation
- 14:28Model performance results
- 16:17Real-time demo
- 17:00Collision prevention detection
- 19:49Industry insights
- 20:32YOLO fine-tuning Q&A
- 21:37Sensor data Q&A
related talks
41mKEYNOTE: Image processing, AI, and autonomous systems
19mAI Trends in Intelligent Transport Systems
26mAI for Autonomous mobile robots
17mAI/ML demo derby
21mUsing Computer Vision to Solve Interesting Problems
25mDeep learning in deep nets: Helping fish farmers feed the world
01h 02mAI on the Edge
46mML Use Cases at Think Big Analytics