soybeans-noted-augumented-cut

The dataset contains 20,539 images of grains annotated with bounding boxes in YOLOv8 format. These images have been pre-processed and augmented to enhance model training for grain classification and defect detection. The following pre-processing steps were applied to each image:

Auto-orientation of pixel data (with EXIF-orientation stripping)
Resize to 600x600 pixels (stretching)
Grayscale conversion (CRT phosphor)
Auto-contrast via histogram equalization

Additionally, the following augmentation techniques were applied to create three versions of each source image:

50% probability of horizontal flip
50% probability of vertical flip
Random rotation between -15 and +15 degrees
Random brightness adjustment between -15% and +15%

This dataset was created to support computer vision models by providing a diverse range of augmented data for better model generalization and accuracy. The images are in JPG format and are annotated to identify grain defects such as moldy, burnt, pecky, scorched, greenish, and good grains.

Format: YOLOv8 bounding box annotations.
Image Size: Resized to 600x600 pixels.
Number of Images: 20,539 images (with augmentations generating additional versions of each source image).