The main task of this project is to create an application which classifies user inputted pictures of wild mushrooms according to their species, using a machine learning classifier based on a convolutional neural network. In result the application calculates the probability for each species of mushrooms, the image may belongs to.

The project is not about the decision whether a mushroom is edible or not. Instead the application should help to understand the biodiversity of mushrooms in Central European forest areas.

- Do not use the information to classify mushrooms according to their edibility -
- Do not use the information to classify mushrooms according to their edibility -
- Do not use the information to classify mushrooms according to their edibility -
- Do not use the information to classify mushrooms according to their edibility -

• /saved_models - include some saved models (<25 MB) from training phases
• /souce - include source code written in python
• data.rar - collected data of mushroom images (sorted)
• proposal - proposal for the udacity ml project
• report - third report and last submission for the udacity ml project
• report_from_first_submission - second report for the udacity ml project
• report_from_second_submission - first report for the udacity ml project

Author: Michael Kögel
Date: 28th July, 2018

Full dataset will be provided after uploading the full project documentation.

• Mushroom World - Website with database for mushroom images
• funga.fi - Finnish website with database for mushroom species
• Google Images - Used to generate RSS Feeds

• Python 3
• Keras (The Python Deep Learning library)
• scikit-learn (Machine Learning in Python)

• Deep Shrooms Project - Project for classifying mushrooms in edible and not-edible
• Research paper - Paper describing support of acquisition of recognition of mushroom images