Research Article

A Benchmark for Endoluminal Scene Segmentation of Colonoscopy Images

Table 3

FCN8 endoluminal scene semantic segmentation results for different numbers of classes. The results are reported on validation set. In all cases, we selected the model that provided best validation results (with or without class balancing).

Number of classesIoU backgroundIoU polypIoU lumenIoU spec.IoU meanAcc mean

492.0739.3759.5540.5257.8892.48
392.1950.7056.4866.4692.82
296.6356.0776.3596.77