The usecase is built on jetson Nano(4GB dev kit)
The usecase inference was executed and tested on jetson Nano.
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.
--- 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
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.
$ sudo nvpmodel -m 0
$ sudo jetson_clocks
Define the following in the config.json file:
Open the terminal in the location where loiter_dwell_detection.py is present.
run the following command:
$ python3 loiter_dwell_detection.py