ImageNet Classification with Deep Convolutional Neural Networks Implementation

1.0 About

A simple implementation of the paper ImageNet Classification with Deep Convolutional Neural Networks using pytorch.

1.1 Model Architecture

2.0 Sample Implementation

2.1 Installation

Install the required packages

pip3 install -r requirements.txt

2.2 ImageNet Data

This project is aimed to use the classic ImageNet Dataset. But due to my limited resources, I will just simply use the Mini-ImageNet Dataset I found in Kaggle.

2.3 Train Dataset

Download and upack the dataset and change the imagenet_data_dir_train variable into the dataset's directory in config.py.

imagenet_data_dir_train = '\ImageNet-Mini\train'

2.3 Trigger Training

To trigger training, simply input the following command in your terminal:

python3 train.py --epochs=100 --batch_size=64 --lr=0.001 --val_split=0.2

Or you can just edit the parameters in variables in config.py and simply use:

python3 train.py

2.5 Test Images

Testing in this project is very simple. You can use the following command for testing where --model_path is the path of your pretrained model and --num_images is the number of random images from your test dataset:

python3 test.py --model_path=mymodel.pt --num_images=10 --print=True

Similar to training, you can either input the --data_path in your run command or simply set your test dataset's directory into the imagenet_data_dir_test variable into the dataset's directory in [config.py].