I'm Michael. Same demo I posted a few weeks ago, no server this time: photo to 3D Gaussian splat, all in the browser.
Depth Anything V2 Small runs via ONNX Runtime Web / @huggingface/transformers. It tries to use WebGPU (which does inference in about 1 second on an M4 Pro processor and 2 seconds on an iPhone 17 Pro) and falls back to WASM (about 5 seconds inference on the same devices).
It's a web component, two lines to embed.
Demo: https://mukba.ng/image-3d/demos/local/ Docs: https://mukba.ng/image-3d/ Source: https://github.com/imichaelnorris/image-3d
Comments URL: https://news.ycombinator.com/item?id=48442524
Points: 2
# Comments: 1