kokjianyu/alexnet

Implementation of AlexNet in Pytorch

Alexnet Implementation

This project is an implementation of the Alexnet that was introduced in the paper "ImageNet Classification with Deep Convolutional Neural Networks" Link to paper

Prerequisite

Required libraries can be found in requirements.txt.

Install with the following command: pip install -r requirements.txt

Dataset

Originally, AlexNet was tested using ImageNet 2012 dataset.

However, usage of this dataset requires approval. As such, I will be testing my implemetation against CIFAR100 instead. However, as the base size of an image from CIFAR (32 x 32) is very different from ImageNet (469x387 on average), my results might differ greatly from the paper.

Code to download the dataset is included in main.ipynb

Training and Evaluation

The code to train and test the implementation is included in main.ipynb.

The model is trained for 50 epochs, and have achieved an test error rate (Top-1) of 49%. This is understandably different from the result achieved by the paper (37.5%) due to difference in training time and dataset.

A pretrained model was not included in the repo due to size limitation of github.