Semantic Segmentation with SegNet

FCN ResNet18 - MHP - 512 x320 the Pre - Trained Segmentation Models to test the effect is not obvious, only color a little dark


Do you use the similar input source as shared in this Github:

You may need to fine-turn the model if the use case have something different.

You might want to increase the alpha value by running the segnet program with --alpha=200 or some higher value. The default value is --alpha=120. Increasing it will make the effect more noticable.

Thank you for your reply, I used the model provided by the official website。
Original image of semantic segmentation input:

Semantic segmentation output picture:

Is this result correct?

You probably want to test the MHP model first on the humans_*.jpg images (i.e. images/human_0.jpg). The MHP models are fairly low-res, so they may have trouble picking up the smaller people in the city image that you tried. You could try that image with Cityscapes model though.