Fast 3D Convolution algorithms Lecture
Nowadays, digital videos are everywhere and revolutionize very many domains, notably:
-Digital Media (video/movie) Content Production and Broadcasting,
-Social Media Streaming and Analytics (g., YouTube),
-Mobile computing and streaming
-Videoconferencing
-Medical/Biological/Dental Imaging and Diagnosis,
-Big Visual Data Analytics,
-Internet and Communications (media broadcasting, streaming).
-Scientific Imaging of any sort, e.g., Physics.
Furthermore, Video Processing and Analysis enables diverse applications, in unison with Computer Vision and Machine Learning:
-Autonomous Systems (cars, drones, vessels) Perception,
-Robotics Perception and Control,
-Intelligent Human-Machine Interaction,
-Anthropocentric (human-centered) Computing,
-Smart Cities/Buildings and Assisted living.
Visual Computing, encompassing Computer Vision and Video Processing and Analysis, coupled with AI (notably Machine Learning and Deep Neural Network) advances hit the news almost every day.
This lecture overviews Fast 3D Convolution algorithms that has many applications in the fast implementation of 3D image and video filtering, 3D CNNs and motion estimation. It covers the following topics in detail: 3D linear and cyclic convolutions, Fast 3D convolutions by using FFTs, Block-based methods, Optimal Winograd 3D convolutions.