Image-Classification

This repository is a PyTorch implementation of an image classifier based on the LeNet-5 CNN architecture.

Notes:

Tested for python >= 3.6
Only tested for CPU

Table of Contents:

Install
Run
Data
Results

Install

# clone repo
pip install -r requirements.txt

Run

Use python main.py to run the preset configuration to train and evaluate the model. The preset configuration can be found in hyperparams.ini.

To run a custom experiment use python main.py <experiment name> <params>. For example:

python main.py -n test_fashion_1 -d fashion -b 32 --lr 0.0001

You can evaluate a pre-trained model with the following:

python main.py -n test_fashion_1 --is-eval-only

Output

Running will create a directory results/<saving-name>/ which contains:

model.pt: The trained model.
specs.json: The parameters used to run the program (default and those modified with the CLI)

Help

usage: main.py [-h] [-d {mnist,fashion}] [-b BATCH_SIZE] [--lr LR] [-e EPOCHS]
               [-m {Lenet5}] [-n NAME] [-s SEED] [--is-eval-only] [--no-test]

PyTorch implementation of convolutional neural network for image
classification

optional arguments:
  -h, --help            show this help message and exit

Training specific options:
  -d, --dataset {mnist,fashion}
                        Path to training data. (default: fashion)
  -b, --batch-size BATCH_SIZE
                        Batch size for training. (default: 64)
  --lr LR               Learning rate. (default: 0.0005)
  -e, --epochs EPOCHS   Maximum number of epochs to run for. (default: 15)

Model specific options:
  -m, --model-type {Lenet5}
                        Type of encoder to use. (default: LeNet5)

General options:
  -n, --name NAME       Name of the model for storing and loading purposes.
                        (default: fashion_1)

Evaluation specific options:
  --is-eval-only        Whether to only evaluate using precomputed model
                        `name`. (default: False)
  --no-test             Whether or not to compute the test losses.` (default:
                        False)

Data

Current datasets that can be used (these will download by themselves):

Adding your data

To use your own data you should package it as a PyTorch dataset constructor. For more information see the link:

dataset

You can do this by adding your data to the YourData class at the bottom of the file datasets.py.

Results

Pre-trained models for fashionMNIST and MNIST can be found in the results folder.

The following results were achieved on the fashionMNIST dataset:

Epochs: 15
learning rate: 5e-4
batch_size: 64

***************************************************
*            Evaluating Train Accuracy            *
***************************************************

Train accuracy of the network on the 60000 test images: 98 %

Accuracy of T-shirt/top : 96 %
Accuracy of Trouser : 100 %
Accuracy of Pullover : 95 %
Accuracy of Dress : 98 %
Accuracy of  Coat : 98 %
Accuracy of Sandal : 99 %
Accuracy of Shirt : 95 %
Accuracy of Sneaker : 99 %
Accuracy of   Bag : 99 %
Accuracy of Ankle boot : 96 %

***************************************************
*            Evaluating Test Accuracy             *
***************************************************

Test accuracy of the network on the 10000 test images: 91 %

Accuracy of T-shirt/top : 81 %
Accuracy of Trouser : 100 %
Accuracy of Pullover : 81 %
Accuracy of Dress : 95 %
Accuracy of  Coat : 88 %
Accuracy of Sandal : 97 %
Accuracy of Shirt : 78 %
Accuracy of Sneaker : 97 %
Accuracy of   Bag : 96 %
Accuracy of Ankle boot : 92 %

And for the MNIST dataset:

Epochs: 10
learning rate: 5e-4
batch_size: 64

***************************************************
*            Evaluating Train Accuracy            *
***************************************************

Train accuracy of the network on the 60000 test images: 99 %

Accuracy of 0 - zero : 100 %
Accuracy of 1 - one : 100 %
Accuracy of 2 - two : 99 %
Accuracy of 3 - three : 99 %
Accuracy of 4 - four : 99 %
Accuracy of 5 - five : 99 %
Accuracy of 6 - six : 100 %
Accuracy of 7 - seven : 99 %
Accuracy of 8 - eight : 99 %
Accuracy of 9 - nine : 99 %

***************************************************
*            Evaluating Test Accuracy             *
***************************************************

Test accuracy of the network on the 10000 test images: 99 %

Accuracy of 0 - zero : 100 %
Accuracy of 1 - one : 100 %
Accuracy of 2 - two : 99 %
Accuracy of 3 - three : 98 %
Accuracy of 4 - four : 98 %
Accuracy of 5 - five : 100 %
Accuracy of 6 - six : 100 %
Accuracy of 7 - seven : 99 %
Accuracy of 8 - eight : 99 %
Accuracy of 9 - nine : 98 %

linesd / Image-Classification