Korean dataset has upper categories such as rice/soup/noodles/meat side dishes/vegetables side dishes/fish side dishes/kimchi
And there are 53 subcategories, and the data have 100~150 images per category
The 600 images were increased to 2,400 images using data augmentation.
Data augmentation was performed using Roboflow, and optimal results were obtained when Flip, 90 Rotate, Crop, Rotation, Shear, Saturation, Brightness, Exposure, and Blur were used.
The example program using this data is as follows.
KoreanFood recognition using yolov5