MCC-Net: A class attention-enhanced multi-scale model for internal structure segmentation of rice seedling stem

•A non-destructive segmentation method for the internal structure of rice seedling stems based on X-ray CT and deep learning was presented.•A class attention-enhanced multi-scale model was proposed to segment internal structures of different sizes for rice seedling stems.•The coordinate spatial attention (CSA) module was designed to automatically locate the target area.•The internal structure areas of different sizes for rice seedling stems were accurately segmented.