This work examines how a forced-attention technique can be applied to the task of Video Activity Recognition. The Look & Learn system performs early fusion of critical detected areas of attention with the original raw image data for training a system for video activity recognition, specifically the task of Squat “Quality” Detection. Look & Learn is compared to previous work, USquat, and achieved a 98.96% accuracy on average compared to the USquat system which achieved 93.75% accuracy demonstrating the improvement that can be gained by Look & Learn’s forced-attention technique. Look & Learn is deployed in an Android Application for proof of concept and results presented.
KEYWORDS: Machine vision, Sensors, Video processing, Machine learning, Computer vision technology, Neural networks, RGB color model, Data modeling, Systems modeling
Recently, Exercise Analysis has gained strong interest in the sport industry including athletes and coaches to understand and improve performance, as well as preventing injuries caused by incorrect exercise form. This work describes a system, USquat, that utilizes computer vision and machine learning for understanding and analyzing of a particular exercise, squatting, as proof of concept for the creation of a detective and corrective exercise assistant. Squatting was chosen as it is a complicated form of exercise and is often mis performed. For USquat, a Recurrent Neural Network is designed using Convolutional Neural Networks and Long Term Short Term networks. A sizable video library dataset containing numerous “bad” forms of squatting was created and curated to train the USquat system. When evaluated on test data, USquat achieved 90% accuracy on average. On a developed Android application that uses the resulting model, upon detection of “bad” squatting forms, it offers an instructive “good” video related specifically to the user’s bad form. Results including live application outcomes are demonstrated as well as challenging video results, problems, and areas of future work. Early work on the creation of a follow-on system to USquat that automatically generates a custom video of the user performing a correct action for the purpose of learning proper activity performance. Additionally, early work on a different version of USquat that explores an attention mechanism network is discussed.
In times of health crises disease situation awareness is critical in the prevention and containment of the disease. One indicator for the development of many contagious diseases is the presence of fever and the proposed system, IRFIS, extends prior research into fever detection via infrared imaging in two key ways. Firstly, the system utilizes a modern, machine learning based object detection model for detecting heads, supplanting the traditional methods that relied upon shape matching. Secondly, IRFIS is capable of running from the Android mobile platform using a small, commercial-grade infrared camera. IRFIS’s head detection model when evaluated on a dataset of unseen images, achieved an AP of 96.7% with an IoU of 0.50 and an AR of 75.7% averaged over IoU values between 0.50 and 0.95. IRFIS calculates the target’s maximum temperature in the detected head sub-image and real results are presented as well as avenues of future work are explored.
Responding to health crises requires the deployment of accurate and timely situation awareness. Understanding the location of geographical risk factors could assist in preventing the spread of contagious diseases and the system developed, Covid ID, is an attempt to solve this problem through the crowd sourcing of machine learning sensor-based health related detection reports. Specifically, Covid ID uses mobile-based Computer Vision and Machine Learning with a multi-faceted approach to understanding potential risks related to Mask Detection, Crowd Density Estimation, Social Distancing Analysis, and IR Fever Detection. Both visible-spectrum and LWIR images are used. Real results for all modules are presented along with the developed Android Application and supporting backend.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.