summitkwan

5 ms·

1. 1.▲
   by summitkwan 8y ago

   Thanks for mentioning this! I think other than using this model for enhancing Photoshop, Instagram, etc. Police sketching would be an awesome application of it
2. 2.▲
   by summitkwan 8y ago

   Thanks for the reply! I just want to add a little more details. As long as the training dataset is diverse enough, the trained GAN generator should be capable of generating images that are similar to the training dataset. Therefore there
3. 3.▲
   by summitkwan 8y ago

   That would be super cool idea and I have thought about it! To deal with music, we may want to use RNN, LSTM or WaveNet instead of GAN as the backbone generator and use my tl-GAN idea to find the "feature" axes for controlled gene
4. 4.▲
   by summitkwan 8y ago

   Thank you for the clarification! The latent space is so well behaved that that I can find linear relationship between the latent vector and the feature labels, and then use the regression slop as the "understanding" of the space
5. 5.▲
   by summitkwan 8y ago

   Nice try, you managed to push to the limit of this algorithm!
6. 6.▲
   by summitkwan 8y ago

   Hi, I am the author of this blog post, feel free to ask me any questions about this work! Thank you for your comments!