This project is an implementation of Cutout, Mixup and Cutmix on CIFAR100 dataset. We explore the overfitting rate of each data augmentation
- Python 3.8
- Download this program through git clone and put it in your repository
- You can refer to this environment setting
cd your_direction
conda create --name neuralnetwork --clone base
conda activate neuralnetwork
conda install pytorch torchvision torchaudio cudatoolkit=10.2 -c pytorch
conda install -c conda-forge tensorboardx
- Run the training code directly to train the models
cd direction/baseline/
python main.py
cd direction/baseline_mixup/
python main_mixup.py
cd direction/baseline_cutout/
python main_Cutout.py
cd direction/baseline_cutmix/
python main_CutMix.py
- Run the code of analysis and tensorboard results
cd direction
python analysis.py
tensorboard --logdir=tensorboard_direction --bind_all