🧠 Most latent-space demos stop at a scatter plot. This one lets you click a point in the UMAP atlas, watch a chair resolve from a 256-dim code, then scrub a slider and see it melt into a lamp, meshed live from a neural SDF. Point clouds, geodesic paths, 64³ to 128³, all in WebGL.
🔗 webgpu.com/showcase/latentsd…#WebGL#ThreeJS#DeepSDF#MachineLearning#CreativeCoding
I'm obsessed with this, I've been dreaming of it since I first saw a t-SNE map of the principal components of DeepSDF. Really want to work on doing this for stable audio latents to make a new age customizable theremin.
I originally developed this tool for MSc students and early PhD students approaching the field. Hope you find this helpful!
Kudos to the DeepSDF authors: arxiv.org/pdf/1901.05103.pdf
Excited about Neural Fields and 3D Computer Vision? :) Check out this user-friendly DeepSDF repository: github.com/maurock/DeepSDF/. Unlike other (excellent) resources, you can install it with a single command and use it for both reconstruction and shape completion!
DeepSDF is one of the initial works that eventually led to the proposal of NeRFs for rendering 3D geometries. It is incredibly simple to understand, and (in my opinion) a great building block in knowledge for deeply understanding the NeRF framework. Check out my overview below!
Join our 3D Deep Learning reading group to discuss exciting topics like Neural Fields (NeRF, DeepSDF, OccNet, etc.) and their applications in both virtual and physical environments. Please visit 3d-deeplearning-rg.github.io… to sign up!