yc015/scene-representation-diffusion-model
Linear probe found representations of scene attributes in a text-to-image diffusion model
This project helps researchers and artists understand and manipulate how text-to-image models create scenes. By adjusting the model's internal representation of elements like foreground objects, you can guide it to generate a series of images that show an object moving, without needing to retrain the model. This is useful for anyone exploring the capabilities of generative AI for creative content or studying model behavior.
No commits in the last 6 months.
Use this if you want to create short video clips with moving foreground objects from a single text prompt using an existing text-to-image model, or if you're a researcher exploring how these models represent and control scene attributes.
Not ideal if you're looking for a simple, out-of-the-box tool for general video generation or if you want to fine-tune a model for specific styles.
Stars
35
Forks
6
Language
Jupyter Notebook
License
MIT
Category
Last pushed
Jul 11, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/yc015/scene-representation-diffusion-model"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
OPTML-Group/Unlearn-Saliency
[ICLR24 (Spotlight)] "SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in...
Shilin-LU/VINE
[ICLR 2025] "Robust Watermarking Using Generative Priors Against Image Editing: From...
WindVChen/DiffAttack
An unrestricted attack based on diffusion models that can achieve both good transferability and...
koninik/DiffusionPen
Official PyTorch Implementation of "DiffusionPen: Towards Controlling the Style of Handwritten...
Wuyxin/DISC
(ICML 2023) Discover and Cure: Concept-aware Mitigation of Spurious Correlation