George Profenza

You can see the prompt in the ComfyUI screenshot.

This is a rough screenshot of the dataset

Another digital human / digital avatar test, this time using a LORA trained Wan2.1 model on 27 images. (I've used ComfyUI for inference/tests. This is WIP: just 14 epochs trained) #wan21 #machinelearning #diffusion #comfyui #avatar

The dataset was generated using VHAP ( github.com/ShenhanQian/VHAP ). (The mesh is about 10K triangles)

The source was a monocular 15 seconds clips shot on my 4 year old phone

First tests trying out Gaussian Avatars ! Congratulations to the teams at @TU_Muenchen and @Woven_Toyota for the amazing work ! project: shenhanqian.github.io/gaussi… paper: arxiv.org/abs/2312.02069 code: github.com/ShenhanQian/Gauss… #avatar #gaussiansplatting #machinelearning

The @ChiliPeppers logo from a math formula as mesh plot (using @GnuOctave) 🤓😎

I've been fortunate enough to make it to the @ChiliPeppers concert this weekend. The solos from @johnfrusciante were blissful. This is my nerdy #mathart tribute using @GnuOctave

Also got to see some tasty dithering that made the qualities of the original medium shine through. Artworks by the very talented @Andy_Macgregor

Bonus points: moving the acetate stencil over the print results in a nice moiré optical illusion spiral animation. (No #processing nor computer required 😌)