Skip to content
/ cifar Public

Evaluating different convolutional neural networks capabilities on the CIFAR-10 dataset

Notifications You must be signed in to change notification settings

viadanna/cifar

Repository files navigation

MLND Capstone Project

by Paulo Viadanna

This projects uses deep learning to identify objects in the CIFAR-10 dataset. As such, it'll use the libraries scikit-learn, keras, tensorflow, pandas and numpy. For plotting, matplotlib and seaborn will be used. Check the requirements.txt file for further information.

Folder structure:

  • experiment.py is the main entry point that runs the models.
  • models.py contains each model implementation using Keras.
  • preprocessing.py is a helper to download, preprocess and augment the dataset.
  • inception_v3.py contains the pre-trained Inception model, not used but kept for historical reasons.
  • inception_hack.py contains the tweaked Inception model that was used here.
  • run_all.sh is a simple script to run the models as specified in the report.
  • results_analysis.ipynb is a Jupyter notebook used to generate the plots.
  • helpers/ folder contains simple scripts to import training and results data.
  • input/ will contain the preprocessed datasets.
  • output/ will store the training history and results for each model.

About

Evaluating different convolutional neural networks capabilities on the CIFAR-10 dataset

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages