This project is supposed to do hand written numbers (0 ~ 9) recognition by using a softmax learning model without any intermediate layers. The data set we used for training and testing is Mnist. Then, a cross entropy loss function is applied to evaluate the difference between the predictions and the correct labels. The loss is minimized by using gradient descent method.
- Tensorflow installed
- Python Version: 3.5 installed
- Github installed
Download this repository:
git clone https://github.com/jb892/OCR_LR.git
python ocr_lr_test.py
Average accuracy of the trainned network on testing dataset = ~ 92%
This project is licensed under the MIT License