Quantifying biotic damage on plants, by separating plant-images and subsequently operating a convolutional neural network

A computer-implemented method (602B, 702B) to quantify biotic damage in leaves (140, 142) of crop-plants (110, 112), the method (602B, 702B) comprising:

• receiving (601A) a plant-image (410/412) showing a crop-plant (112), the plant-image (412) showing the aerial part of the plant (112), with at least the stem (120, 420), branches (130, 430) and leaves (140, 440) and showing the ground (150, 450) underneath the plant (112), on that the plant (112) is placed;
• processing the plant-image (412) by a segmenter module (262) to obtain a segmented plant-image (462) being a contiguous set of pixels that shows a contour (460A) of the aerial part, the contour (460A) having a margin region (458) that shows the ground (150, 450) partially;
• and
• processing the segmented plant-image (462) by a convolutional neural network (272) that uses regression to obtain a damage degree, the convolutional neural network (272) having been trained by processing damage-annotated segmented plant-images (461).