Skill LevelArea of FocusOperating SystemPlatform/Hardware
IntermediateComputer Vision, IoT, Smart CitiesLinuxDragonBoard 410c

Our project is an attention monitoring and alert system, that is designed to use Deep Neural Networks to detect person(s) not paying attention and creates visual and audio alerts. Using the DragonBoard™ 410c from Arrow Electronics, the system also aims to record hourly data to generate a graph that can be used to assess “alertness” during certain tasks like office work, driving, handling critical equipment, etc. It accommodates for normal body movements and minor obstructions with automatic pausing based on person(s) in view.

The inspiration behind this project was the need for a personalized system that could monitor attention, performance, and rest. Also, this technology could be used as an in-car system to monitor the driver’s focus.

We hope to create a product that consumers can implement at their work stations to monitor their focus and assist them with suggestions for taking breaks. These suggestions would be based on historical data that shows when optimal times to take breaks are. Also, by using a development platform like the DragonBoard 410c, we hope to make our system portable and be used in places without computers, like automobiles.

DragonBoard 410c Project setup

Example of a focused personExample of an unfocused person

Distractions count graph

The system is implemented using Python2.7 on a Debian operating system.

Launch the python application using the following command,

python --prototxt deploy.prototxt.txt --model res10_300x300_ssd_iter_140000.caffemodel

Hit q on keyboard to stop the program.

Shakti Sumanshaktisuman01@gmail.comInterim Engineering InternQualcomm Technologies, Inc.
Dhvani Pateldhvanipatel227@gmail.comInterim Engineering InternQualcomm Technologies, Inc.
Nicholas Neighborsnick.neighbors2@gmail.comInterim Engineering InternQualcomm Technologies, Inc.
Balakrishnan ThipirajapuramVenkataramanbalame1600@gmail.comInterim Engineering InternQualcomm Technologies, Inc.