Skip to main content

Table 8 Classification results on unpolished rice when ResNet-101 is used as backbone in the weakly supervised Grad-CAM approach

From: Deep learning based high-throughput phenotyping of chalkiness in rice exposed to high night temperature

ResNet-101 Acc.(%) Chalky Non-chalky
Pre.(%) Rec.(%) F1(%) Pre.(%) Rec.(%) F1(%)
Polished 63.01 0.00 0.00 0.00 63.01 100.00 77.31
Unpolished 83.43 98.50 82.19 89.61 43.65 91.67 59.14
Mixed 84.20 98.08 83.45 90.18 44.77 89.17 59.61
  1. Three models are evaluated: 1) polished model trained on polished rice images; 2) unpolished model trained on Unpolished (12); 3) mixed model, obtained by further training the polished model using the Unpolished (12) images. Performance is reported in terms of Accuracy (Acc.), Precision (Pre.), Recall (Rec.) and F1 measure (F1). Precision, Recall and F1 measure values are reported separately for the Chalky and Non-Chalky classes. All three models are evaluated on the test subset corresponding to the Unpolished (12) rice images. The best performance for each type of model for each metric is highlighted using bold font