Recent Advances in Online Object Tracking

Explore recent advancements in online object tracking through this comprehensive guest lecture by Dr. Ming-Hsuan Yang at the University of Central Florida. Delve into various tracking approaches, including sparsity-based classifiers, discriminative models, and occlusion handling techniques. Examine the collaborative model, qualitative and quantitative evaluations, and tracking by detection methods. Learn about compressive tracking, Gaussian PDF assumptions, and experimental results. Gain insights into evaluation issues, methodologies, and datasets used in tracking algorithms. Discover temporal robustness evaluation techniques and one-pass evaluation methods for low-resolution scenarios.

Intro
Yosemite National Park
Lake Tahoe
Lake Mono and Parker Lake
Tracking Approaches
Related Work
Sparsity-based Classifier
Training Data
Discriminative Model: Summary
Occlusion Handling
New Histogram
Collaborative Model
Qualitative Evaluation
Quantitative Evaluation
Concluding Remarks
Outline
Tracking by Detection
Algorithm Overview
Two Components
Revisit MILTracker
Constructing Random Matrix R
Compressive Tracking/Sensing?
JL vs. RIP
Gaussian PDF Assumption
Experimental Results
Motivation
Evaluation Issues
Tracking Algorithms
Evaluated Algorithms
Evaluation Dataset
Evaluation Methodology
Temporal Robustness Evaluation
One Pass Evaluation
Low Resolution