This example shows the neural network (multilayer perceptrons) to paint an image
- paint_tensorflow.py (implemented with Tensorflow)
- paint_keras.py (implemented with Keras)
My inspiration: http://cs.stanford.edu/people/karpathy/convnetjs/demo/image_regression.html
- 7 hidden layers.
- Input layer have 2 ( x, y coordinate or pixels on an image).
- Output layer have 3 (red, breen and blue).
Input to the model are x, y coordinates (on an target image).
** In python, size of an image is height x width
If there is one input to the model.
If there are many inputs (many examples) to the model.
For target dataset to train my model, I use real an image as the target dataset. When neural network learns from the target image, it can predict 3 outputs are red, green, blue (RGB).
The picture on the left is original and the right hand side is created by my neural network.