Committed to Further Testing of V7 Darwin Image Labelling for Machine Learning at Scale

image
image
image

Today, after testing several image labelling tools, we committed to do an extensive labelling on V7 Darwin to label the thousands of pictures we took at Hoi Ha Wan. We've particularly liked the active learning and "Magic Wand" function to select complex shapes such as corals. But wait, what is V7?

And why do we need V7?

To train our AI to classify corals with our coral reef mapping robot. Our corals have very complex shapes, drawing the outline would take a very long time. V7 makes our job much faster and accurate.

I found this very nice poster made by Hong Kong AFCD to help us cheat sheets to classify corals:

image

Currently, we only use 10 classes:

image