Image Classification

Classifying flowers, plants and mushrooms!

I started out by scraping the names of individual species of flowers, plants and mushrooms from different websites.

I then scraped the web for images of each species and built separate models to classify a large number of species within each category.

As a last step before training, I used Isolation Forest for anomaly detection, i.e. to find images that were outliers with regards to the target domain.

I used transfer learning methods by leveraging large pretrained models and fine-tuning them to my datasets. In addition I tested a number of data augmentation techniques to arrive at the best possible model.

The current best mushroom model is deployed as a simple Gradio app with which users can upload images to be classified.

My goal is to build a mobile app that can be used in the wild!

Find it here

Tools and techniques

• Python
• Fast.ai / PyTorch
• Gradio
• Image classification
• Anomaly detection
• Web scraping