MushPy Project - Mushroom Recognition

Content

Here enclosed are all files related to the MushPy project which includes:

• Data collection & data cleaning
• Exploratory data analysis & data visualization
• Models
• Report

All files are in French language.

Context

MushPy project is a Data Science project carried out as part of the Data Scientist training at DataScientest. It has the aim to allow the students to be entirely involved in a Data Science project in the conditions of a professional data team. This project is 120 hours long minimum over 3 months.

The objective of the MushPy project is to create recognition models of mushrooms thanks to the computer vision. Indeed mushrooms have a large number of species, genuses, families, as well as a wide variety of biological and nutritional characteristics. It may be interesting to know how to classify them precisely using computer vision. The enclosed report to this repository shows the work done to reach this goal, starting with data collection, data cleaning, data exploration and data visualization, the application of the different models and finally the results obtained.

Result

The best result was achieved with an accuracy of 77% thanks to Learning Transfer based on "EfficientNetB1". This result has been achieved to classify 5 families of mushrooms.

Project Team

• Sevil CARON
• Adrien MOREAU
• Fernando GONÇALVES
• Thibault KACZMAREK