WeedCrop Image Dataset

Data Description

• Total Images: 2,822
• Annotations: YOLO v5 PyTorch format
• Pre-processing:
  • Auto-orientation of pixel data (EXIF-orientation stripping)
• Augmentation:
  • 90-degree rotations (none, clockwise, counter-clockwise) with equal probability
  • Random shear: -15° to +15° horizontally and vertically
  • Random brightness adjustment: -25% to +25%

Classes

• Crop
• Weed

Augmentation Details

To enhance the dataset and improve model robustness, each source image undergoes the following augmentations to create three versions:

• Rotation: Images are rotated by 0°, 90°, or 270°.
• Shear: Random horizontal and vertical shear between -15° and +15°.
• Brightness: Random adjustment between -25% and +25%.

Inspiration

Accurate identification and differentiation of weeds from crops are crucial for effective farming practices, enabling better crop management and yield optimization.

Applications:

• Precision Agriculture: Helps in the development of automated systems for weed detection and removal, reducing the reliance on manual labor and herbicides.
• Research and Development: Facilitates research in machine learning and computer vision for agricultural applications, fostering innovation in the sector.
• Educational Purposes: Serves as a valuable resource for academic projects and training machine learning models in agricultural studies.