Currently Ai-based applications are available, but none of these offers a guarantee of truthfulness. Since in nature there are several chameleon mushrooms, the application may not be able to recognize it.
The idea of this project offers a more accurate solution for the mushroom recognition based on 3 different images to uniquely validate the mushroom: Cap, Stem and Gills of Mushroom.
The project involves the application of image recognition techniques based on deepLearning algorithms, for now trained with caffe framework.
Once the prototype environment for the workflow has been developed, specific test and validation datasets will be made for the most common mushrooms.
Furthermore, the use of Movidius in the prototyping phase is envisaged, which uses the training models realized with the help of the colfax cluster to do inference from the edges without the cloud.