Skip to main content
. 2020 Jul 8;20(14):3816. doi: 10.3390/s20143816

Figure 8.

Figure 8

Example segmentation results on the TACO test set. Input images and ground-truth annotations are shown in the first two columns. Baseline methods are FCN-8s (VGG-16) and DeepLabv3 (ResNet-101). Our proposed methods (FCN-8s-ML and DeepLabv3-ML) more accurately recover object boundaries. Best viewed electronically, zoomed in.