- We had many different ideas because the field of machine learning is extremely big and are various possibility to do a project. At first, we thought of mixing ML5 with a GAN Model to have an interactive input, but it was more complicated as we expected, and it also did not fit our dataset. Helena found the Lunar and Planetary Laboratory website from the University of Arizona, where images of Mars have been published every few days since 2006.
- We scraped the images with Webscraber.io and cropped them in the right size (512x512).
I experimented with a small set of 64 photos and there have been trained with the ADA GAN 2 notebook. At the same time we had trained on Paperspace and Jupyter with the same notebook a bigger data set of 28´895 photos. We made some research what we could do further with the trained model.
The Animation that Derrick Schultz did with a trained Model of Flowers, we found exciting and we wanted to try out with the pictures from the mars. We found a notebook in Derrick Schultz's repository called flesh digressions which fitted well with the output we were looking for. We all three experimented with the flesh digressions Notebook with different trained Models also with Pre-Trained Model and various settings. The animations are very varied from colorful to monochrome depending on the settings and the checkpoints of the model.
The technical part and Python were a bit challenging, but we learned a lot about how machine learning works, even if many parts are still like a black box.
MODUL + TASK:
Machine learning
Briefing: Try out different approaches for Machine Learning
General Topic: Machine Learning
WHEN + DURATION:
5th Semester Digital Ideation
SOFTWARE + TOOLS:
Webscraper.io, Photoshop, Python, Colab, Paperspace, Jupyter
FURTHER INFORMATION
_ Teamwork 3 Pers.
My part was to do the Pre-Training with Colab ADA GAN2, Animations with Colab Notebook flesh digressions
more details project on GitHub
MODUL + TASK:
Virtual Reality
Topic: Virtual Reality
Briefing: Design an aspect of a laboratory in the near future. Build a virtual dashboard to monitor operational performance of the lab
equipment.
-Interaction with the dashboard is done by gesture / hand tracking
-A professional visual appearance of the UI is expected
WHEN + DURATION:
5th Semester Digital Ideation, 2 weeks
SOFTWARE + TOOLS:
Unity, Ocolus Cinema 4D
FURTHER INFORMATION
_ Teamwork 3 Pers.
My part was research, design, Cinema 4D Visualiszation, integrate to Unity
_ distance learning due to covid 19
