Project
Sail-CV is an open-source computer vision framework that makes sail analysis quantitative and automated. It combines two modules designed for embedded hardware: tell-tale tracking (detection, tracking, and state estimation from a single onboard camera) and 3D sail reconstruction (metric point clouds from calibrated stereo cameras, without markers or applied texture). The goal is a practical, reusable base for sailors, engineers, and researchers—real-time tell-tale analytics and metric sail geometry from the same pipeline.
Genesis
In 2025 I was a performance and data engineer for Team France (America's Cup). There, sail shape was routinely measured with very expensive LiDAR. I wanted to challenge that with a more accessible approach—and to use the same hardware to obtain tell-tale states for flow attachment. That is the genesis of Sail-CV: one vision stack for both shape and tell-tales, open and adaptable.
Publication
An article presenting this framework has been submitted for INNOVSAIL 2026.
Links
- Code: Sail-CV (core) · Sail-CV-MLops (training and MLOps)
- Models & data: Weights · Dataset (Hugging Face)
- Author: If you want to know who I am, you can find my personal web page here.
Related work
Sail-CV builds on foundation models for 3D vision (e.g. MASt3R) and real-time detection (RT-DETR). Related sailing and aerodynamic work includes tell-tale and tuft analysis (e.g. Dhome et al., Innov’Sail; Lee et al., RAL). See the full article and references for details.