In this repo, image segmentation is tested using Mask RCNN. If an image is input, the model will segmentize into 90 classes. The class names can be checked in the mscoco_labels.names The model's result is still required to fine-tune on a specific dataset. You can download model file here
@Citation https://github.com/computervisioneng/semantic-segmentation-tensorflow-opencv