Colab code: link
Project folder: link
Working with images requires very careful analysis. There are a lot of things that can go wrong, and today we will be covering the idea of duplicates. I know it might be tempting to go over the images one by one if you have a fairly small dataset. However, let’s be honest — it’s quite a tedious and prone-to-error approach.
This notebook provides a complete solution that tackles the following tasks:
- Gather and clean data
- Create and remove duplicates both automatically and manually
- Train CNN model with ResNet50
- Analyze misclassifications