Plastic Classification using image processing through OpenCV's DNN module utilizing a Yolov3 model.

First, make sure the latest version of Python is installed.

Create a virtual enviornment in the project folder. The final folder in the path should be named Env.

python -m venv C:\Path\To\Virtual\Environment

Activate the virtual environment and install dependencies:

C:\Path\To\Env\Scripts\activate
(Env) pip install -r requirements.txt

Add the necessary files into the resources folder. The .cfg and .names files are included, But the .weights file will need to be downloaded.

To run this program first activate the virtual environment and then run the Plastic_Classifier_GUI.py file.

C:\Path\To\Env\Scripts\activate
(Env) python Plastic_Classifier_GUI.py

The output file gives a label and 4 values, Center X, Center Y, Width, and Height. To find the bounding boxes using these values, do as follows :

1. Subtract Width/2 from X and Height/2 from Y to get the bottom left corner of the box.
2. Add Width to X and Height to Y to get the top right corner of the box.

From plasticClassifier.py:

#Lines 109 and 110. Values are later rounded in lines 52 and 53.
x = center_x - w / 2
y = center_y - h / 2

#Lines 54 and 55
x_plus_w = round(x+w)
y_plus_h = round(y+h)