Simple CNN with PyTorch

Description

This repository is based on one of the university homework assignments on data visualization and serves to describe the process of building a classification model in a simple computer vision task.

Context

The topic of the assignment is classification of cells for finding malaria cases. This dataset is taken as a basis.

All code has been split into separate files for readability. The project has the following structure:

• The file assignment.ipynb contains the progress of solving the assigned task,
• The cnn package defines the model itself and auxiliary constants, functions and classes for working with it and data,
• train.py provides all the necessary functionality to train the model,
• model.pt contains the final machine learning model,
• The logs folder contains logs of training of individual models in the process of hyperparameter optimization,
• The data folder stores the dataset,
• extra contains everything that does not fit into the project structure defined above.

Priority in commenting is given to visualizations rather than to the machine learning algorithms themselves.

Installation

git clone https://github.com/tejpaper/simple-cnn.git
cd simple-cnn
pip install -r requirements.txt

References

1. Data source
2. Feature maps visualization tool

License

MIT