Fig. 1From: Weakly supervised segmentation models as explainable radiological classifiers for lung tumour detection on CT imagesWeakly supervised Unet and standard CNN architectures. Blue layers represent inputs and outputs. Red, green and yellow layers represent downsampling, upsampling, and non-resampling convolutional blocks, respectively. Purple, charcoal and lilac layers represent convolutional, dropout and normalisation layers, respectively. Orange layers represent dense layersBack to article page