CNN-CIFAR-10-Image-Classification

Simple CNN for CIFAR-10 image classification task using Pytorch

Dataset

CIFAR-10

CNN

PCA

Reduce to 50 components by scikit-learn PCA, plot first two components.

t-SNE

Further reduce to two dimension by t-SNE in sklearn.

Result

92.8% accuracy after 30 epochs.

Run

• Install Anaconda
• Create a conda env that contain python 3.7.5: conda create -n your_env_name python=3.7.5
• Activate the environment (do this every time you open a new terminal): conda activate your_env_name
• Install the requirements into this conda env: pip install --user --requirement requirements.txt
• Run the jupyter notebook: jupyter notebook

Reference

• Pytorch
• scikit-learn