Charlotte's
Handwritten Digit Recognizer
Powered by a neural network implemented with keras and tensorflow.js
| DATA: | 70 000 images of size 28 x 28 with 3 channels consisting of: | |
|
42 000 images Provided by the mnist database 
|
28 000 images of generated data
|
|
| MODEL: |
Convolutional Neural Network: Keras Sequential Model with 6 hidden layers  Optimizer: Adam Learning Rate: 0.001 Loss: Binary Crossentropy |
| EVALUATION: | |
   
|
 |