https://github.com/gentaiscool/cnn-autoencoder-tf
CNN and Contrastive Autoencoder (CAE) on EMNIST using Tensorflow
https://github.com/gentaiscool/cnn-autoencoder-tf
autoencoder cae cnn contrastive emnist tensorflow
Last synced: 3 months ago
JSON representation
CNN and Contrastive Autoencoder (CAE) on EMNIST using Tensorflow
- Host: GitHub
- URL: https://github.com/gentaiscool/cnn-autoencoder-tf
- Owner: gentaiscool
- License: mit
- Created: 2018-06-01T06:11:43.000Z (about 8 years ago)
- Default Branch: master
- Last Pushed: 2018-10-07T12:16:02.000Z (over 7 years ago)
- Last Synced: 2025-03-28T22:11:23.539Z (about 1 year ago)
- Topics: autoencoder, cae, cnn, contrastive, emnist, tensorflow
- Language: Python
- Homepage:
- Size: 96 MB
- Stars: 10
- Watchers: 3
- Forks: 6
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Convolutional Neural Network and Contrastive Autoencoder on EMNIST
In this project, we are going to evaluate the performance of convolutional neural network (CNN) and contrastive autoencoder (CAE) models by conducting empirical study on simple image data (EMNIST dataset) [1]. This dataset consists of 28x28 images of handwritten characters that belong to 47 classes.
[1] Gregory Cohen, Saeed Afshar, Jonathan Tapson, and Andre van Schaik. EMNIST: an
extension of MNIST to handwritten letters. arXiv preprint arXiv:1702.05373, 2017.
### Run the code
#### Package required
- Python 3.5 (or later)
- Tensorflow (https://www.tensorflow.org/)
#### Parameters
- lr: initial learning rate
- mm: momentum
- bsz: batch size
### CNN
#### Run the code
Train CNN
```
python --task="train_cnn" --lr=0.1 --mm=0.2 --bsz=32
```
Cross Validation CNN
```
python --task="cross_valid_cnn"
```
Test CNN
```
python --task="test_cnn" --lr=0.1 --mm=0.2 --bsz=32
```
### Autoencoder
#### Run the code
Train AE
```
python --task="train_ae" --lr=0.1 --mm=0.2 --bsz=32
```
Cross Validation AE
```
python --task="cross_valid_ae"
```
Test AE
```
python --task="evaluate_ae" --lr=0.1 --mm=0.2 --bsz=32
```
### Note
COMP5212 - Machine Learning Programming Assignment 2 in HKUST