Train an neural network to identify digits written on a canvas
In this project I focused on helping others understand how AIs work by allowing visitors to train an fully untrained neural Network.
This is a Neural Network build using TensorFlow JS Library.
| Layer (type) | Output shape | Param # |
|---|---|---|
| input1 (InputLayer) | [null,28,28,1] | 0 |
| conv2d_Conv2D1 (Conv2D) | [null,27,27,64] | 320 |
| re_lu_ReLU1 (ReLU) | [null,27,27,64] | 0 |
| max_pooling2d_MaxPooling2D1 | [null,13,13,64] | 0 |
| conv2d_Conv2D2 (Conv2D) | [null,12,12,64] | 16448 |
| re_lu_ReLU2 (ReLU) | [null,12,12,64] | 0 |
| max_pooling2d_MaxPooling2D2 | [null,6,6,64] | 0 |
| flatten_Flatten1 (Flatten) | [null,2304] | 0 |
| dense_Dense1 (Dense) | [null,300] | 691500 |
| dense_Dense2 (Dense) | [null,100] | 30100 |
| dense_Dense3 (Dense) | [null,10] | 1010 |
Total params: 739378
Trainable params: 739378
Non-trainable params: 0