Finding the Dots: Why Colour Detection Failed and Subtraction Worked

Deep in the Wild, part 4. Third of six technical posts on recovering machine-readable bird annotations from eleven years of aerial survey screenshots. By the end of the last post the pipeline could re