https://github.com/loggerhead/nnocr
A simple English characters recognizer using 3 layers neural network.
https://github.com/loggerhead/nnocr
Last synced: 3 months ago
JSON representation
A simple English characters recognizer using 3 layers neural network.
- Host: GitHub
- URL: https://github.com/loggerhead/nnocr
- Owner: loggerhead
- License: wtfpl
- Created: 2016-01-16T15:03:28.000Z (over 9 years ago)
- Default Branch: master
- Last Pushed: 2017-03-17T03:35:32.000Z (over 8 years ago)
- Last Synced: 2025-01-26T06:09:04.933Z (5 months ago)
- Language: Matlab
- Size: 314 KB
- Stars: 0
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# nnOCR
A simple English characters recognizer using 3 layers neural network.
# Installation
```bash
# Install Octave
sudo apt-get install -y octave
./install_octave_packages.sh
# Install python dependences
sudo apt-get install -y python-dev python-numpy python-scipy
sudo pip install -r requirements.txt
```
# Running
## Training
Put your training images to `nnOCR/trainer/training_dataset/` for trainning. The image should:
* Named as *label*\_*identify*.bmp. *label* is the content of image, which is used to classify.
* Size is 20*x*20 pixels
```bash
cd trainer
./main.m [iter_num] [lambda]
```
## Predict
You can start a web server to predict user uploaded images.
```bash
# start http server at `localhost:5000`
python server.py
```
Or you can just running it locally.
```bash
python predict.py IMAGE_PATH
```
NOTICE: The built-in `thetas.mat` can only used to recognize printing numbers.
## Test
```bash
python predict.py test/1.jpg
python predict.py test/2.jpg
```
# Contributors
* [wooyiuhan](https://github.com/wooyiuhan)