You will work with image datasets of rice samples provided in three processing states:

• Paddy rice
• Brown rice
• White rice

Each image is complemented by laboratory-measured reference data on quality generated by industry-standard rice analysis equipment. These measurements serve as the ground truth labels for training and evaluation.

Target variables:

Your model should extract key quality characteristics from images, focusing on:

Grain Count, Structure & Size

• Count
• Broken_Count
• Long_Count
• Medium_Count

Grain Colour

• Black_Count
• Chalky_Count
• Red_Count
• Yellow_Count
• Green_Count

Average Whole Kernel Length to Width Ratio

• WK_Length_Average
• WK_Width_Average
• WK_LW_Ratio_Average

Colour Space Features (CIELAB)

• Average_L
• Average_a
• Average_b

The images are available on Google Cloud: https://storage.googleapis.com/unido-afririce/