Description:

The usecase is built on jetson Nano(4GB dev kit)
The usecase inference was executed and tested on jetson Nano.

This usecase is for generating alert:

1. when loitering is detected in a given area.
2. if count of person increases beyond allowed number.

SSD Model is used for detecting people.

Project Structure:

loitering_dwell_detection

--- 1. loiter_dwell_detection.py
--- 2. config.json 
--- 3. model
    ------ MobileNetSSD_deploy.caffemodel
    ------ MobileNetSSD_deploy.prototxt.txt
--- 4. tracker
    ------ centroidtracker.py
--- 5. person_count_alert
    ------ stores images with time stamp when person exceed beyond threshold value

Parameters and values for generating alert are defined in config.json

config.json containes the following parameters:

1. person_duration (Dwell time allowed per person)
2. personCountExceed (maximum number of persons allowed in a frame at a given time)
3. vid_source (video source[videofile, rtsp, camera])

Note: To use Opencv with CUDA and use opencv dnn module install Opencv from source.

Before running the script run the following commands to maximize the device preformance:

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

Steps to run the the script:

1. Define the following in the config.json file:
   

   1. video source
      
   2. person count allowed
      
   3. dwell time allowed
      
      

2. Open the terminal in the location where loiter_dwell_detection.py is present.
   

3. run the following command:

   $ python3 loiter_dwell_detection.py