Skip to main content

Table 2 Scores of models with E-loss based on modified architectures (except GapNet)

From: Multi-labelled proteins recognition for high-throughput microscopy images using deep convolutional neural networks

Model with E-loss Macro F1
ResNet50 (fold 0) 0.770
ResNet50 (fold 1) 0.766
ResNet50 (fold 2) 0.768
ResNet50 (fold 3) 0.758
ResNet50 (fold 4) 0.762
InceptionV4 (single fold) 0.767
SENet (single fold) 0.782
Dense50 (random fold 1) 0.758
Dense50 (random fold 2) 0.761
Dense50 (random fold 3) 0.766
ResNet101 (single fold) 0.778
GapNet-PL 0.765
  1. The bold of “0.782” means that the modified SENet performs best in this stage