niklastiede / imageclassification

Image classification model of the FashionMNIST dataset. A little project to learn pytorch and DL.

Image Classification - PyTorch/FashionMNIST

The model predicts categorical data (10 categories of clothing) using the FashionMNIST dataset. Achieved accuracy with the present CNN is 92.7%.

• The train.py script trains the weights of the model based on the given hyperparameters
• tensorboard is used to compare the yielded accuracies of different hyperparameter optimizations
• the eval_conf_matrix.py script creates a confusion matrix to evaluate the trained model in more detail (how well each image category was classified)

This project helped me getting familiar with pytorch and concepts of DL. It relies on the work of deeplizard.