Here are two tensorflow implementation GANs applications on Emoji.
Emoji WGANs(improved) in the directory named 'emoji_wgan_gp' for learning a mapping from input uniform distribution to output Emoji images , for example:
Emoji Pix2Pix in the directory named 'emoji_pix2pix' for learning a mapping from input sketch image ( black-and-white painting ) to output Emoji images , for example:
cv2
tensorflow
scipy
numpy
codes are typed in jupyter notebook which is convenient to debug codes .
Besides the jupyter notebook , some tool code are included. The utils.py can help you visualize the images in a grid , and the canny.py can help you get the sketch images from original images.
On the training process , the loss information will be saved in the logs directory . using tensorboard can see the loss about the Generator and Discriminator.
Code refer heavily to Pokemon_GAN , which give me some inspiration to implement the application on Emoji.