fabiocigaina/TACO-yolo11s

TACO YOLO11s

This model performs object detection to identify litter in images.

Links

• Demo: https://huggingface.co/spaces/fabiocigaina/TACO_YOLO_Demo
• Code: https://github.com/FabioCigaina/YOLO-Litter-detection

Base model

Ultralytics YOLO11s (yolo11s)

Dataset

• TACO Dataset: http://tacodataset.org/
• Kaggle mirror: https://www.kaggle.com/datasets/sohamchaudhari2004/taco-trash-detection-dataset

Preprocessing:

• Selected the 9 most frequent supercategories
• Grouped Unlabeled litter and the rest into the Other class
• Id to class:

  Id	Class
  0	Plastic bag & wrapper
  1	Cigarette
  2	Bottle
  3	Bottle cap
  4	Can
  5	Other plastic
  6	Carton
  7	Cup
  8	Other

Training

• 120 epochs
• imgsz: 896
• mosaic: 0.5

Evaluation metrics

Metric	Score
mAP50	0.311
mAP50-95	0.234