Overview / Usage

Different industries usually have a queue problem. In the manufacturing industry, it could be to count the number of people in an assembly line. For retail, it could be to count the number of people in a checkout terminal etc.

This work can be used by sending the feed from a camera to a computer (with Intel's CPU, IGPU, Vision Unit, and/or FGPA) which has this application.

Not only does it recognizes queues, it will also count the number of people in each queue and display the result.

For the retail scenario, once the number of people in a queue is more than 5, it redirects people to other less busy queues. It does the same thing for the manufacturing scenario, albeit for 6 people. For the transport scenario, the number of people in a queue have to be 15 before it redirects them.

Methodology / Approach

A person detection model from Intel's model zoo is used to detect people. For three different scenarios i.e. manufacturing, retail, and transportation, different model precisions and hardware types are selected.

The three different scenarios that depict real-world problems for Computer Vision on the egde are provided in "Project Scenarios.pdf."

An app is then built to integrate the model. The performance of the model for the three different scenarios are simulated using Intel's DevCloud.

The manufacturing, retail and transportation notebooks submit the jobs to Intel's DevCloud to run the inferencing jobs on various Intel's Hardware and produce a simulated result on the performance metric (frames per second, model load time, and inference speed) of the system on various Intel hardware.

The create_python_script notebook writes out the smart_queue detection app, while the create_job_submission script is a pipeline to submit various inferencing jobs from the given case studies.

The hardware proposal and a comparison of the performance of the various hardware is given in Choosing the Right Hardware.pdf.

Technologies Used

Intel OpenVINO

Intel DevCloud

Python

OpenCV

Intel FGPA

Intel Movidius NCS

Intel Core i5 6500 TE

Intel IGPU

Documents and Presentations

Repository

https://github.com/ajudges/smart_queue/