This repository contains PyTorch classification model based on Pneumonia X-Ray image dataset.
This model was based on CNN publication by PhD Krzysztof Sopyla: PyTorch CNN Tutorial CIFAR-10.
Dataset is available here: https://www.kaggle.com/paultimothymooney/chest-xray-pneumonia
Model description:
1. Data download process, unziping and clearing useless data.
2. Libraries import, definition of required classes and methods.
3. Samples import, files compability verification and new samples generation process based on imgaug library:
In this section preprocessing of the images were performed. Basic dataset contains almost 6000 pictures. Number of images was multiplied to ~25000 by sequential operations based on imgaug library.
Training pictures visualisation:
4. Parameters definition, images to tensors transformation, and model definition:
Images size were set on 224x224 with 1 channel, batch_size was equal 16, two output classes called: 'NORMAL' and 'PNEUMONIA'. Model was trained on 20 epochs.
5. Model Summary and training process:
6. Training results visualisation:
If there is problem to render .ipynb file please check: https://nbviewer.jupyter.org/github/JMcsLk/PneumoniaXRayPyTorch/blob/master/PyTorch_X_Rays_Conv2D.ipynb