PiP Package for converting the masks for grayscale to color.
There are a number of public datasets avalilable for testing your semantic segementatio models. Each data set is unique and has its own challenages, problems and complexity level. Each data represents different scenarios of real life like autonomous driving, person body part, common objects etc.
Usually the list of calss names and particularly the color pallet is not provided wiht the dataset so I collected the names of classes and their corresponding color pallets and added in this repo. you can use them to color your results.
Simply put semantic segmentation is a dense classification problem in which the CNN are supposed to classify each pixel of the input image w.r.t its corresponding class.
Some of the popular datasets are given below you can also follow the link to download them (you might have to register wiht your work email first.)
If you wanna do panoptic (semantic + instance) segmenation, then you will also need a list of stuff and things classes like in case of vistas and cityscape so I have also provided them inside the scripts. Although a lot of semantic segmentation datasets also have their panoptic varients like cityscape, vistas etc. Here I'd like to mention the one I have worked on for panoptic segmentation.