Assignment 9
After slightly struggling with the Python and RNN, I found Runway much easier to use, as it has pre-trained models and the interface intuitive. Hopefully, I’ll get better at using Runway and use it for my final project!
As for this week’s assignment, I looked at StyleGan, which generates photorealistic images of faces, landscapes, and more. Searching for ways to how to use it, I came across the CodingTrain tutorial for using StyleGan.
Following the tutorial, I chose Cupcakes as my checkpoint and after running it, I was given a gallery of different cupcakes. I picked my favorite one:
I successfully ran it on my p5 web editor to generate different cupcake images whenever I press the “cupcake’ button. One shortcoming I realized is that the p5 won’t run or generate images if Runway is not running (it stops automatically??), just like how it works with p5 serial communication with Arduino. I have to continuously go back and forth to see if Runway is running.
This exercise, although it doesn’t have a lot of creative elements, was very helpful for me to familiarize myself with Runway.
I honestly prefer using Runway to ml5.js and python as I think there is a lot of space for creativity given that there are already pre-trained models. Overall, the terms used in the software are more straightforward (e.g. truncation, HTTP, etc.) than the terms used in neutral network such as epoch and normalizeData.