Box Detection in Warehouse Conveyor Belts

This Repo contains a Box Detection Application capable of identifying box containers in conveyor belt pictures.

This project has many applications in Warehouse Inventory Management:

• Accurate Counting: Automatically count boxes on conveyor belts for real-time inventory updates, reducing manual counting errors and delays.
• Box Type Identification: Differentiate between different types of boxes based on size, shape, or markings, enabling efficient sorting and tracking.
• Damage Detection: Identify damaged boxes during transportation, allowing for quicker intervention and quality control.

End User Instructions to run the Object Detection Model

In order to use this object detection web application, please follow below instructions

1. Install Docker
   • Make sure Docker is installed on your system. You can download and install Docker Desktop from the official Docker website.
2. Pull the Docker Image
   • Open docker desktop, followed by a terminal or command prompt in your system and run:

     docker pull agpsuai23/box_detection_image

   • This command will download the Docker image named agpsuai23/box_detection_image to your local machine.
3. Run the Docker Container
   • Once the image is pulled, execute the following command:

     docker run -p 8000:8000 agpsuai23/box_detection_image

   • This command starts the container and maps port 8000 of the host to port 8000 of the container, allowing access to the application.
4. Access the Application
   • Open a web browser and go to: http://localhost:8000/
   • You can now interact with the Box Detection Application through the browser interface.

Project Details

1. Computer Vision based Data Science & ML
   • Yolo8_Model_Creation.ipynb downloads the dataset from Roboflow, followed by training Yolov8 on train set using transfer learning, validation and test result display. We end up with a mean Average Precision (mAP) of 0.913 (90%+). Finally, the model is saved as best.pt
2. Software Engineering using FAST API
   • I created a FastAPI web application (main.py) to serve the saved model to customers.
   • Home page (Upload_for_Detection.html) is a basic I/O based UI that lets a user upload a conveyor belt image, and when a button is clicked, the application uses the previously saved Yolo8 model to detect box containers and return the image with labels in a new page. Code can be found in main.py - @app.post("/YOLO_Box_Prediction_Website/")
   • Alternatively, I also created an API functionality that can programmatically accept a conveyor belt image as input to return as output the image with labels saved in local system. Working can be found at main.py - @app.post("/YOLO_Box_Prediction_Service/"). An example of API call can be found in CUSTOM_YOLO_API_call.ipynb
3. Deployment using Docker Containerization
   • Finally, I deployed this FastAPI based Yolov8 application as a Docker Image (based on Dockerfile) and uploaded it to Docker Hub.
   • All they need to do is to simply install Docker Desktop, pull the image from the hub, run it as a Docker container and finally access the webpage on a browser as localhost:8000
   • Advantage of containerization is that all end users can avoid the hassle of installing packages mentioned in requirements.txt, which are necessary to run the application on their system.