/Dog-breed-NN

:dog: Experimenting with neural networks to classify images of 5 different dog breeds

Primary LanguagePython

Dog breed neural network 🐕

Experimenting with neural networks to classify images of 5 different dog breeds.

Repo for testing how changing metaparameters and setup of a neural network affects the end accuracy.

Classes:

  • Poodle
  • German Shepherd
  • Dalmatian
  • St Bernard
  • Pug

Compatible with:

  • AlexNet
  • VGG-19
  • ResNet-15
  • ResNet-34
  • ResNet-152

Generates training & validation graphs over the training period. Creates a confusion matrix of results.

Before training, run split.py to split the dataset into the desired training / validation / test split.

To run from command line:

  • --mod specify which model
  • --lr specify the learning rate
  • --a specify augmentations (1 for on, 0 for off)
  • --s specify the datasplit