Generative AI — Course Assignments

Overview

Two compact assignments covering implicit representations and denoising diffusion.

• A1 – Implicit Fields: From point clouds to surfaces via an MLP decoder (occupancy/SDF-style training).
• A2 – Diffusion (2D): Train a noise-prediction network and sample denoised points on toy datasets.
• **One Paper presentation about 3D Diffusion Automated Creation (Craftsman)

Topics Covered

• Implicit neural representations (occupancy/SDF)
• MLP decoders, activations, regularization
• Loss design and training dynamics
• Diffusion basics (forward noising / reverse denoising)
• Noise schedules, sampling, and evaluation

run assignment scripts/notebooks (examples)

python assignments/implicit_fields/train.py python assignments/diffusion/run_sampling.py