ASA ML GAMEPLAY - TO BEE OR NOT TO BEE

All pictures used to train all neuronal network described here have been supplied by :
  • Philippe Moniotte for the Eristalis ones, Those Eristalis are mimetics from Apis
  • The INRA Team in Avignon through two research projects namely Florapis and Florabeilles for the apis ones.
Testing test provided here comes from NL and US websites classified data for Eristalis, from the Florapis project for Apis
RUN
Our Object detection model follows the ResNet50 convolutional Faster-RCNN architecture.

It allows to locate in the image the bugs (bees or flies) and allocate them to a class:
  • Apis = melliferous bees
  • Eristalis = Eristalis (Tenax or Arbustorum) One of the mimetic for those bees

The model was trained on 70 images of melliferous bees and 70 images of two Eristalis species.