This dataset contains 3 classes of content annotations in YOLO format:
overlay_captions: 0
overlay_graphics: 1
overlay_text: 2
Training set: 436 images
Validation set: 109 images
The dataset follows the YOLO format:
dataset/
├── images/
│ ├── train/
│ └── val/
├── labels/
│ ├── train/
│ └── val/
└── dataset.yaml
Each label file contains annotations in the format:
class_id x_center y_center width height
