Creating an AR Shopping Experience Leveraging Real-Time Neural Rendering
Learn how to use the newest SnapML model on real-time neural rendering to create powerful AR shopping experiences! Lens Studio can now display realistic and detailed geometry, using only 2D images. Making AR Shopping experiences amazingly realistic through NeRF, a rendering technique that uses machine learning to take a limited set of input images and display 3D objects from any viewpoint.
In this tutorial we walk through the following components:
00:00 Introduction
1:10 Volumetric rendering vs polygonal rendering
2:07 MobileR2L - a new approach to rendering
2:43 How to set up MobileR2L data in Lens Studio
3:15 How the dataset can be used
5:27 Sample Lenses with Snapcodes
5:58 How a model is trained
Lens Studio template you can use as reference: https://github.com/Snapchat/snapml-te...
Supporting Materials
Github Neural Rendering folder: https://github.com/Snapchat/snapml-te...
Inference Notebook: https://github.com/Snapchat/snapml-te...
Collect the Dataset and train from scratch one: https://github.com/Snapchat/snapml-te...
Lens Studio project examples: https://github.com/Snapchat/snapml-te...
Visit ar.snap.com/learn to learn more about Lens Studio
Visit docs.snap.com for technical documentation
Visit ar.snap.com to download Lens Studio