I am curious, what software was used for labelling the images in the U-Net segmentation and Mobile Net programming assingments.
The images may come from outwith deeplearning.ai, but curious, what software or libraries would be used in practice for similar problems to get the training dataset in the first place.
It’s a really good question, but I do not know the answer. This falls under the general rubric of “Data Science”, which is a separate discipline from Machine Learning. As Prof Ng comments at numerous points, the data here is fundamental, but in these courses he doesn’t really discuss how to create and curate labelled datasets.
The dataset for the U-net assignment is from the CARLA self driving vehicle project. One thing to do would be to take a look at their website, but I tried doing that a while back and did not really find much information about the dataset.
The other approach would be to just google “how to create data for image segmentation” and see what you find. I tried a variant of that and found this medium article but it looks like it might just be a slightly disguised advertisement for a company that provides this service. Maybe you can find better links. 