Fourier Feature Networks and Neural Volume Rendering

Channel:
Microsoft Research
Subscribers:
344,000
Published on ● Video Link: https://www.youtube.com/watch?v=Al6NTbgka1o


Duration: 26:13
4,302 views
211

Fourier Feature Networks are an exciting new development in Computer Vision, and their use for modeling radiance fields has produced a range of impressive results at the meeting point of Computer Vision and Computer Graphics. In this lecture, Matthew covers the motivation behind using Fourier features in neural network training, introduces the fundamentals of volumetric ray casting, and then shows how we can use Fourier Feature Networks to render high-quality novel views of complex 3D scenes.

This guest lecture was part of University of Cambridge Engineering Tripos Part IIB, 4F12: Computer Vision.

Speaker: Matthew Johnson, Research Software Engineer Lead at Microsoft Research, Cambridge
https://www.microsoft.com/en-us/research/people/matjoh/

Access code and a Jupyter notebook for use in further learning at the companion Github repository: https://github.com/matajoh/fourier_feature_nets



Other Videos By Microsoft Research


2022-01-04Talk: What Makes Multi-modal Learning Better than Single (Provably)
2022-01-04Inauguration Ceremony of MSR Asia Theory Center Opening Speech from Zhi-Ming Ma
2022-01-04Plenary Talk: Theory in an industry lab
2022-01-04Plenary Talk: From mathematics to data science and back
2022-01-04Talk: Causal inference, observational studies, and the 2021 Nobel Prize in Economics
2022-01-04AI for Programming Education
2022-01-04Talk: Recent results on stochastic 3D Navier-Stokes equations
2022-01-04Talk: Theoretical Aspects of Gradient Methods in Deep Learning
2022-01-04Recap video of 2021 MSR Asia Theory Workshop (Long version)
2022-01-04SysSieve: Extracting Actionable Insights from Unstructured Text
2022-01-03Fourier Feature Networks and Neural Volume Rendering
2021-12-20DIABLo: a Deep Individual-Agnostic Binaural Localizer
2021-12-16Our Genomes, Our Selves?
2021-12-01A law of robustness and the importance of overparametrization in deep learning
2021-11-29Research for Industries (RFI) Lecture Series: Steven K. Frey
2021-11-24Research for Industries (RFI) Lecture Series: Matthew Realff & Christopher Jones
2021-11-23On Race and Technoculture
2021-11-08Acrylic, metal & a means of preparation: Imagining & living Black life beyond the surveillance state
2021-11-01FastNeRF: High-Fidelity Neural Rendering at 200FPS [Extended]
2021-10-29Full-Body Motion from a Single Head-Mounted Device: Generating SMPL Poses from Partial Observations
2021-10-29Litmus Predictor


Tags:
Fourier Feature Networks
Computer Vision
modeling radiance fields
neural network training
Microsoft Research
Matthew Johnson