This project is a demo that performs a detection of the closest person in a real-time video. The calculation chosen to evaluate who is closer to the camera is the person's height (Change the calculation to your context of use). A value of -1 means that no person is detected.

This value is written to a MongoDB collection evey frame (given by OpenCV).

This project also contains code with a simple Flask app to make requests to obtain the value

1. Pull the code
2. Create a python environment (e.g. Virtualenv), activate it and install de requirements

e.g.

virtualenv -p python3 .venv

source .venv/bin/activate

pip install -r requirements.txt

3. Install MongoDB on your machine and run it as a service (e.g. through Homebrew on macOS)

e.g. With Homebrew

brew services start [email protected]

4. Check that everything works by executing the mongo command

5. Since the code in this project works with specific collections, either create them or modify the code (and create them as well)

e.g. Inside the mongo environment, execute

use ssexpo

db.data

db.data.insertOne({"description":"User proximity value", "value":100000})

db.data.find()

These commands create the DB, the collection, inserts the document we'll be working and retrieves all documents in the collection

6. Everything ready!!

In this project, there are two separated parts, the person-detection part and the Flask part, so the steps I follow to 'run' the application are the following:

1. Make sure MongoDB is running as a service (e.g. brew services list to check)
2. Launch the flask server (this will also launch the recognition process/window)

python flask_server.py

If everything works, navigating to http://127.0.0.1:5000/closest_person on a browser should give you a numeric value (the value stored in MongoDB)

Note for Windows users: Take into account that it might be neccessary to change the Python executable in flask_server.py to the one used in the virtual environment (e.g. .venv\Scripts\python)

3. Launch the person detection script

python detect_and_write_to_mongo.py --prototxt MobileNetSSD_deploy.prototxt.txt --model MobileNetSSD_deploy.caffemodel

A small window should open, showing the input from your webcam

3. To validate that everything works, you can either refresh the web browser or perform commands through the mongo terminal

This gif shows an example, retrieving the value from Unity

• mongo_test.py: A simple test I did to learn MongoEngine
• real_time_object_detection.py: If you run it, the model will detect all the object it's been trained for (saves nothing to MongoDB)
• real_time_people_detection.py: Only detects people (saves nothing to MongoDB)
• real_time_closest_person_detection.py: Filters to get the closest person (saves nothing to MongoDB)

• Real time object detection tutorial: https://www.pyimagesearch.com/2017/09/18/real-time-object-detection-with-deep-learning-and-opencv/
• MongoDB Python tutorial: https://realpython.com/introduction-to-mongodb-and-python/
• MongoDB Community Edition installation: https://docs.mongodb.com/manual/administration/install-community/
• MongoEngine (MongoDB ORM for Python): https://docs.mongoengine.org/tutorial.html