Explore visual material recognition in this guest presentation by Dr. Ko Nishino from the University of Central Florida. Delve into generative and discriminative approaches for recognizing materials visually. Learn about novel reflectance models, radiometric decomposition, and single image material estimation techniques. Discover how to tame material appearance through visual material traits and convolutional material trait kernels. Examine material category classification, failure cases, and the role of materials as visual context. Gain insights into the latest advancements and challenges in the field of visual material recognition over the course of this hour-long lecture.
Overview
Syllabus
Intro
Acknowledgements
Visually Recognizing Materials
Generative vs. Discriminative
Generative Material Recognition
Taming Reflectance
A Novel Reflectance Model . An expressive yet parametric (isotropic) BRDF model
Connecting the Slices
The Space of Real-World BRDFs
Radiometric Decomposition
Single Image Single Material Estimation
A Reflectance Prior
Single Material Results
Single Image Multimaterial Estimation
Multimaterial Results
Using Multiple Images
Reflectance as A Bandpass Filter
Natural Illumination Entropy Prior
Reflectance and Natural Illumination from A Single Image
Real Object / Real Illumination
Real Images
Discriminative Material Recognition
Taming Material Appearance
Visual Material Traits
Convolutional Material Trait Kernels
Material Trait Features
Material Trait Recognition
Materials from Material Traits
Material Category Classification
Material Category Failure Cases
Material as Visual Context
Summary - Towards visual material recognition
Taught by
UCF CRCV
