Head Detection in Public Transit Hotspots

Table of contents

• About
• Features
• Packages Used
• Setup
• Usage
• Screenshots
• References

About

This use case deals with the identification and counting of heads in video footage to understand the Footfall of people in a given area and automatically count the number of people entering and exiting a given location. The current implementation of the use case is specifically focused on automated ticketing systems where the number of people entering or exiting a given vehicle is monitored.

Features

• The project utilizes Single Shot based Detectors (SSD) for Head Identification from video frames.
• The People Counting Functionality is implemented by:
  • Region of Interest Identification
  • Centroid Tracking and head Tagging

Packages Used

• TensorFlow - 1.15.4
• Keras - 2.2.4
• FastAPI
• OpenCV
• imutils
• nest-asyncio
• pyngrok
• asgiref

Setup

Requirements

• Python 3.6+

Installation

• To install required packages:
  

pip install -r requirements.txt

Usage

python main.py

Screenshots

Video Upload UI

Preview Uploaded Video

View Detected Heads and Total Count

References

• https://github.com/AVAuco/ssd_head_keras
• https://github.com/shijieS/people-counting-dataset
• https://www.pyimagesearch.com/2018/07/23/simple-object-tracking-with-opencv/