Pitcher counts/area measurements #2
Description
Ideally we should do full semantic segmentation for individual pitchers, then calculate average count and area per plant and treatment. This seems infeasible with classical methods. A neural net could probably do it, which requires manually tracing pitchers to label training images. This will take some time.
In the meantime we can consider total area for each plant, and estimate pitcher counts by combining a few heuristics:
- total plant area
- convex hull vertex count