4D Dream

• 

  Project picture

Inspiration

We as team entered AI Artathon 2.0, in the beginning we noticed a dilemma, because almost every project from the other teams seemed fancy yet it was pretrained, thus when we wanted to innovate, we had two choices either coming up with an idea and experimenting and publishing a paper or we could use models in a pipeline to aggregate their tasks and make them more useful, unfortunately we did not qualify in Artathon 2.0.

What it does

it creates the whole elements of a music video, poetry, music, video clip. more on that https://youtu.be/UFmA7bWlP1k

How we built it

we built it using a couple of pre-trained models, the whole details of the process is on the github repository https://github.com/MohammedAlkhrashi/TMA

Challenges we ran into

we ran into an out of space problem, because generating the photos is dependent on number of epochs, the more the number the better the result, yet the bigger the number the bigger the number of pictures (frames) we need to aggregate in a video

Accomplishments that we're proud of

we are proud of this project to be honest, the generation of poetry, then out of poetry the generation of melody and the clip, the idea of this pipeline we are really proud.

What we learned

Artathon 2.0 took four days, and we learned a lot through this period, how to connect models, the problems that could arise from that, how decode what tasks are more important than others..

What's next for 4D Dream

we will keep grinding after the loss, which was a true shock, but we will keep going and we will win soon with this idea.

Built With

• python
• pytorch