https://github.com/shiqimei/digit-recognition
Recognize the handwritten digits online with FCNet which is powered by MNIST dataset 😄
https://github.com/shiqimei/digit-recognition
computer-vision handwritten-character-recognition neural-network visualization
Last synced: over 1 year ago
JSON representation
Recognize the handwritten digits online with FCNet which is powered by MNIST dataset 😄
- Host: GitHub
- URL: https://github.com/shiqimei/digit-recognition
- Owner: shiqimei
- License: mit
- Created: 2019-07-10T17:36:26.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2019-08-31T07:31:27.000Z (almost 7 years ago)
- Last Synced: 2025-02-28T20:35:05.902Z (over 1 year ago)
- Topics: computer-vision, handwritten-character-recognition, neural-network, visualization
- Language: Python
- Homepage: https://handwritten.lolimay.cn
- Size: 26 MB
- Stars: 7
- Watchers: 1
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## Introduction
Let's try to recognize the handwritten digits online with our single hidden layer Fully Connected Neural Network (aka. [FCNet](./server/Training/FCNet.py)), which is powered by the [MNIST](http://yann.lecun.com/exdb/mnist/) dataset 😄.
## Playground
It will be nice if you can play with the project. Below are some basic guidelines for you to have a good start!
### Training & Evaluating
You can run [run_Train_Model.py](./server/Training/run_Train_Model.py) and [run_Evaluate_Model.py](./server/Evaluating/run_Evaluate_Model.py) to train and evaluate the model respectively.
### The Network architecture
Currently, we use a single hidden layer Fully Connected Network and make it be trained with MNIST dataset. Feel free to rdesign your own network architecture and replace the file [FCNet.py](./server/Training/FCNet.py) in the **server/Training** folder.
### Play with the utilities
We prepared some helpful tools for you in the [server/toos](./server/tools). You can do some work for the MNIST dataset with the help of these tools.
## License
This project is open source under the [MIT](./LICENSE) license.