One of CS236G's goals is to prepare you to apply GANs to real world tasks. The final project will get you started in that direction.

Class Project Deliverables

20%, Proposal
  • Inspo: Find sources of inspiration that you think are cool. You must include 7 sources, at least 2 of which must be academic papers, and write a brief paragraph on each. In your paragraph, give a brief mention of the sources that they relied on. You will be assessed on the diversity of material (to prevent student mode collapse :D).
  • Idea spew: Create a list of 2+ project ideas you’d be interested in exploring. In each, you must include the dataset you plan to use, the model architecture, and the real-world application this will be used towards. At least two of the three among {dataset, model, application} must be novel. A novel dataset is one which has never had a GAN applied to it (we encourage you to get creative / scrape the data yourself).
15%, Milestone 1
  • I haz the data: Your dataset will be submitted as part of our course’s “data stash”. We get really psyched when we have a good stash. 💫
  • Inkling: Run a baseline model on your data and get cracking!
15%, Milestone 2
  • Ignition on: You should have largely implemented the plan you set sail for ⛵️ (or an alternative prospect that you pivoted to).
  • Star-gazer: Describe some highly risky experiments that you plan to perform. The more the merrier! 🤩
50%, Final Draft
  • Team Playa: Equal participation within your pair (N/A if flying solo). This to reduce the case where one person does all the work, and make it a bit easier on an individual.
  • Virtually-famous: Submit a video no longer than two minutes about your project.
  • Don Quixote: Give the results of the quixotic experiments that you mentioned in Milestone 2. This will be weighted heavily as we want to encourage fearless exploration - don’t worry if experiments don’t pan out!