Suyash Sonawane , Deep Learning Enthusiast | Google DSC Lead 2020 | Developer, Student, Creator.
25th Aug 2021

Imagine a dance video generated by an AI in which not only new dance steps are generated but the AI also creates it's own body to perform those steps.
Here I made a system which can accomplish this task to a smaller extent, I used 2 Neural Networks a LSTM and a GAN. The LSTMs are used to predict the next step given previous 5 steps in the form of body points, the GAN then generates a frame of image from the points predicted by the LSTM that contains the human figure. The architecture of the Neural Networks is provided in the Github repo and for the GAN I used a special GAN type known as StyleGans



Give a star to encourage!


Be the first one to comment!