My colleagues study plant-pollinator networks. Insect (pollinator) identification is a task that requires a lot of effort.

I would like to know if there are already good enough trained or trainable models to identify pollinators (insects) from mobile phone images taken in the field (possibly, a close-up photo of the insect on a flower). Identification to family level would be good enough for their research. Maybe for some easy to identify insects, identification to specie level would be also possible.

Note that, I could find few articles that provide code (e.g. a GitHub repository with the model and workflows), but they are mostly focused on using sharp, good quality, close-up images taken in the lab from insect collections, with the goal to identify insects to species level, e.g:

  • Valan, 2019 - Automated Taxonomic Identification of Insects with Expert-Level Accuracy Using Effective Feature Transfer from Convolutional Networks https://github.com/valanm/off-the-shelf-insect-identification
  • Buschbacher, 2020- Image-based species identification of wild bees using convolutional neural networks, https://deepabis.github.io/, https://github.com/deepabis/deepabis
  • More Valentin Ștefan's questions See All
    Similar questions and discussions