Fig. 3: Segmentation of cell image library dataset with a LACSS model trained with incomplete annotations.
From: Efficient end-to-end learning for cell segmentation with machine generated weak annotations
a A representative example of segmentation results showing the input image (left), the ground-truth segmentation (middle) and model prediction (right). The pseudo colors were generated via skimage’s label2rgb function for visual clarity and carried no extra meaning. b Model performances quantified by average precisions and false negative rates at various IOU thresholds. Scale bar represents 50 micrometers.
