Rahul-7323/Object-Detection-System-RPi

A system which uses raspberry pi 4 for object detection and an accompanying web app for displaying the data.

Real-Time Object Detection System using Raspberry Pi, TensorFlow, MERN and Socket.IO

About The Project

In this project, a real-time object detection system has been developed that consists of a Raspberry Pi and a webcam for performing object detection along with a backend and a frontend for storing and viewing the data. The backend has been built using Express, Socket.io and MongoDB Atlas. The frontend has been built using React and Tailwind CSS.

The Raspberry Pi uses ssd_mobilenet_v1.tflite for performing object detection on the video stream from the webcam. The results from the object detection is then sent to the backend through a HTTP POST request which then gets stored in mongoDB Atlas. The frontend displays all the events that the Raspberry Pi generated with details such as the objects detected in a frame of the video stream, the image, the probability of the class of the detected objects and so on.

Socket.IO has been used for creating a bidirectional connnection between the backend and the frontend. When the backend recieves a new event from the Raspberry Pi, it updates the events displayed in the frontend in real-time.

Frameworks and Technology Used

• Raspberry Pi and webcam for capturing the video stream and sending the events to the backend.
• TensorFlow for loading and using ssd_mobilenet_v1.tflite model on the captured video stream.
• Express for creating a REST API in the backend.
• MongoDB Atlas for storing the data in the cloud.
• Socket.IO for creating real-time bidirectional connection between backend and frontend.
• React for creating the frontend.
• Tailwindcss for styling the frontend.

Getting Started

Backend and Frontend setup

1. Clone the repository.

   git clone https://github.com/Rahul-7323/Object-Detection-System-RPi.git

2. Create a .env file in the root of the backend folder with the same format as that of .env.example file and replace the dummy value of MONGODB_URI with an URL to a real MongoDB Atlas database. Optionally, change the value of the PORT variable to some other port of you choice.
3. Similarily, create a .env file in the root of the frontend folder with the same format as that of .env.example file and replace the values of VITE_BACKEND_API_URL and VITE_BACKEND_SOCKET_URL variables if you have deployed the backend or changed the port numbers to something else in the backend.
4. Run the below commands in the terminal to start the backend.

   cd backend
   yarn install
   yarn dev

5. Run the below commands in the terminal to start the frontend.

   cd frontend
   yarn install
   yarn dev

Raspberry Pi Setup

1. Clone the repository.

   git clone https://github.com/Rahul-7323/Object-Detection-System-RPi.git

2. Replace the value of url variable in the detect_webcam.py file in the raspberry_pi folder to the URL of the backend.
3. Run the below commands in the terminal of the Raspberry Pi to start the object detection process.

   cd Object-Detection-System-RPi/raspberry_pi
   source setup.sh
   python3 detect_webcam.py --modeldir object_detection_model

Usage

For doubts related to usage of this project, watch the demo video.