The problem
Counting and naming animals in survey photos by hand is slow, and different people count differently.
How it works
- 01Images / video
- 02Detect (YOLOv5)
- 03Classify (Inception V3)
- 04Count
- 05Red List status
What was hard
- The rare species matter most — and have the fewest photos to learn from.
- Light and distance change a lot between field photos.
- Accuracy has to be checked species by species, not as one average.
The result
A working pipeline, registered as a copyright: “Wildlife Conservation and Analysis Using Machine Learning”.