no prediction graphs like peacock segmentation as given in the paper

[fgqile]

    Hello, according to the paper ‘’On MSCOCO, it outperforms ResNet-101 by 3.7% and 3.6% mAPs for object detection and instance segmentation‘’. I downloaded the coco dataset, and training as well as testing, however, the training and testing results only have some data metrics values, such as: loss, etc., as in Figure 1, and no instance segmentation visualization effect graph appears, such as: peacock instance segmentation prediction graph. 

I want to get some segmentation graphs like peacock, but after running your program, there are only data metric values, and no prediction graphs like peacock segmentation as given in the paper, as in Figure 2. can you help me to solve this problem, thank you very much.