Open intvo opened 5 years ago
I think maybe other network architecture would suit better road segmentation tasks. This Pytorch repository has some interesting code, I've worked with it and it is really intuitive. This is a Tensorflow repo for semantic segmentation but it seems like it is deprecated. Remember that MaskRCNN is an instance segmentation model, which means that it combines the object detection and semseg tasks to give a more accurate representation than traditional ways that only give bounding boxes, so it works for classes like people, cars, etc., but not for buildings and road.
I would like to train the model to get the road segmentation, is it possible on this code?