/ML_classification

Repositório para o TCC

Primary LanguageJupyter NotebookMIT LicenseMIT

ML Classification

Implementation of Resnet-RS for training, validation and testing

The Resnet-RS models from: https://github.com/nachiket273/pytorch_resnet_rs

Dataset Folders organization:

folder_organization

How to use:

  • In the "Initial configuration" part, replace the ImageFolder path with the path of your own custom dataset
  • There you can also alter the number of epochs, batch size, k-folds of your preference

Values used for this program:

(Those same values were used both for binary and multiclass classification)

  • Number of folds (k): 10
  • Batch size: 52
  • Dropout rate: 0.25
  • Loss function: CrossEntropyLoss()

Pre-trained Network:

  • Number of epochs: 20

NOT pre-trained Network:

  • Number of epochs: 200