This dataset consists of 927 image files that were labeled using Roboflow. The dataset is in YOLOv8 format. The dataset is divided into train, validation and test. Data replication processes were also applied.

Preprocessing

• Auto-Orient: Applied
• Resize: Stretch to 640x640

Augmentations

• Outputs per training example: 3
• 90° Rotate: Clockwise, Counter-Clockwise
• Saturation: Between -20% and +20%
• Blur: Up to 1.5px