Metofico Webinar: Video Annotation for AI Training

Our behavior recognition AI doesn’t work like most others out there. It doesn’t use key point tracking, so it learns in a totally different way to many available solutions.

To train it, we usually take around 10 short (5-minute) videos, which our collaborators have annotated. But when we say annotate, we don’t mean plotting key points, we mean marking when each behavioral event starts and ends, frame to frame.

We’ll be talking all about this in our upcoming webinar on November 28th, including:

• How our AI actually learns from video
• What good annotation looks like (and why it matters)
• How we build the best model for each unique dataset