Fig. 7From: Deep learning based high-throughput phenotyping of chalkiness in rice exposed to high night temperatureExamples of chalkiness binary masks for four unpolished rice grains. The binary masks obtained from the Grad-CAM heatmaps (with ResNet-101 as backbone) using a threshold \(T=60\%\) are shown form the polished, unpolished and mixed models, respectively, by comparison with the ground truth binary maskBack to article page