Spaces:
Sleeping
Apply for community grant: Academic project (gpu and storage)
Breathing New Life into 3D Assets with Generative Repainting
This demo space complements our paper titled "Breathing New Life into 3D Assets with Generative Repainting", recently accepted at BMVC'23 as an Oral presentation by Tianfu Wang, Menelaos Kanakis, Konrad Schindler, Luc Van Gool, and Anton Obukhov.
It allows the user to paint or repaint a 3D model using text guidance powered by Stable Diffusion and NeRF. The demo implements a cached example of painting a horse asset as a unicorn, also seen in the paper.
Processing user uploads requires an A10G small GPU to complete in 20 min. Persistence is optional, as it saves just 1 minute on retrieving the HF cache of the diffusion pipeline. The demo is powered by Docker and Gradio.
Hi @toshas , we have assigned a gpu to this space. Note that GPU Grants are provided temporarily and might be removed after some time if the usage is very low.
To learn more about GPUs in Spaces, please check out https://huggingface.co/docs/hub/spaces-gpus
BTW, do you really need the persistent storage? I see https://huggingface.co/spaces/toshas/repainting_3d_assets/blob/61a0789ca9ec192b15723de9dd04ff23e9ecb9e8/Dockerfile#L8-L10, so removing it without changing this part would break the Space, but I guess you were using persistent storage so you don't have to wait model/data download while debugging. So, maybe you don't need it any more.
Thanks for the grant! The constraints are, of course, reasonable.
Yes, I added persistent storage to cache diffusion models, saving roughly 1 minute of run time.