ASA ML GAMEPLAY - BEES : CLASSIFICATION

The vast majority of the pictures used to train all neuronal networks described here have been supplied by INRA Team in Avignon through two research projects namely Florapis and Florabeilles.
RUN
Our classification model follows the VGG16 architecture with weights adjustment's of the last layer.

Here, it classifies bee images in two classes:
  • Apis = melliferous bees
  • Bombus = bumblebees

The model was trained on 350 images of melliferous bees, 350 images of bumblebees and reached an average accuracy of 92% (out of 50 images tested).