Live Repetition Counting - CNN Architecture and System for Online Motion Detection

Learn about a novel approach to live repetition counting in video streams through this 29-minute lecture from the University of Central Florida. Explore the motivation behind the system, its overall architecture including CNN design, and the integration of counters. Discover how the system handles start, update, and end counting processes, as well as multiple detectors and online mode counting. Gain insights into ROI detection, dataset creation using synthetic data, CNN training techniques, and benchmark results in online mode. Understand the practical applications and performance of this innovative repetition counting system.

Outline
Motivation
Overall System
CNN architecture
Integrating counter
Outer system
Start counting
Update counting
End counting
Multiple detectors
Online mode counting
ROI detection
Dataset
Synthetic data
Training CNN
Benchmark - Online mode
Results