Garbage Detection Model

Detect garbage-like objects from a webcam, images, folders, or video files using YOLO.

The default model is yolov8n.pt, which is trained on COCO. It can detect common waste-like objects such as bottles, cups, bowls, and food items. For real garbage classes such as plastic_bag, cardboard, or food_waste, train a custom model with your own labeled dataset.

Features

• Webcam, image, folder, and video input
• Class filtering for garbage-like objects
• Recursive folder search with --recursive
• Annotated image outputs
• Custom YOLO training script and dataset config

Install

pip install -r requirements.txt

Run On Webcam

python garbage_detector.py --source 0

Press Q to quit.

Run On An Image

python garbage_detector.py --source path\to\image.jpg

Annotated images are saved in outputs.

Run On A Folder

python garbage_detector.py --source path\to\images

Search inside subfolders too:

python garbage_detector.py --source path\to\images --recursive

Run On A Video

python garbage_detector.py --source path\to\video.mp4

Useful Options

Detect every class in the model:

python garbage_detector.py --source 0 --classes all

Use only selected classes:

python garbage_detector.py --source 0 --classes bottle,cup,bowl

Raise or lower confidence:

python garbage_detector.py --source 0 --confidence 0.5

Train A Custom Garbage Model

Create a YOLO dataset like this:

dataset/
  images/
    train/
    val/
  labels/
    train/
    val/

Each label file must use YOLO format:

class_id x_center y_center width height

Then train:

python train_garbage_model.py --data garbage_dataset.yaml --epochs 50

After training, run detection with the best weights:

python garbage_detector.py --model runs\garbage_detector\weights\best.pt --source 0 --classes all

Dataset Classes

The included garbage_dataset.yaml uses these classes:

0 plastic_bottle
1 glass_bottle
2 can
3 paper
4 cardboard
5 plastic_bag
6 food_waste
7 other_trash

Project Files

garbage_detector.py      # Detect from webcam, image, folder, or video
train_garbage_model.py   # Train a custom garbage detector
garbage_dataset.yaml     # YOLO dataset configuration
requirements.txt         # Python dependencies

Notes

Model weights, training runs, datasets, and output images are ignored by Git so the repository stays lightweight.