Handwritten-Characters-Generation-using-GANs
Handwritten Characters Generation using GANs in TensorFlow
- This is one of the cool projects I did in my internship period. It introduced me to GANs.
- The Dataset can be downloaded from here.
- GANs consists of two deep neural networks called Generator and Discriminator.
- For the Generator network, I used 2 Dense layers followed by 2 Conv2D layers.
- Applied Batch Normalisation to each of these layers.
- For the Discriminator network, the architecture is the reverse of that of the Generator network.
- A random noise is generated first with the dimensions and is fed into the Generator network.
- It's output(fake data) is then fed into the Discriminator network.
- The model is trained for 80 epochs and observed quite a lot of improvement during the training period.
