This project focuses on diagnosing diseases in cassava plants, a vital staple crop in Uganda, using images from a Kaggle dataset. The goal is to classify plants as healthy or affected by one of four diseases. Key features include data preprocessing (normalization, resizing, and balancing classes), CNN-based model training, and addressing challenges like overfitting. This solution aims to help farmers quickly identify crop issues and improve yield sustainability.
**In the 022-explore-dataset we will find**
- Our data needs to be loaded and processed
- We can normalize our data to help our models
- If we have unbalanced classes, our model might favor one class over another
- We can correct this by undersampling, throwing away data in the larger classes
-
**In the 023-multiclass-classification:**
- Once more we preprocessed our data to make it ready for deep learning.
- We built a CNN with 3 convolutional and max pooling layers, followed by flattening and two dense layers.
- We used nn.Sequential to easily build our model's architecture by defining the order of the layers.
- Training the model
- Testing the model