mrgloom/awesome-semantic-segmentation

This repository is a curated reference list for a computer vision task called semantic segmentation, which is the process of labeling every pixel in an image with the category it belongs to. For example, in a street photo, a model might color all pixels showing cars in one shade, pedestrians in another, and roads in a third. It is used heavily in self-driving vehicles, medical imaging, and satellite analysis. The list organizes dozens of published research papers and their corresponding code implementations. Each paper has links to one or more code repositories, with notes on which programming library each one uses, such as PyTorch, Keras, or TensorFlow. Readers can find classic approaches like U-Net, SegNet, DeepLab, and FCN, as well as many newer variants. The list also separates methods by the type of segmentation they perform, including standard scene labeling, instance segmentation (which distinguishes individual objects of the same type), and panoptic segmentation (which combines both). Beyond architectures, the list links to commonly used training datasets organized by domain, covering urban street scenes, medical scans, satellite imagery, and indoor environments. There are also sections listing evaluation tools, helper libraries, and survey papers that give broader overviews of the field. This is a reading and discovery resource, not a software package. There is no code to run directly from this repository. Its value is as a starting point for researchers or engineers who want to understand what approaches exist, find a tested implementation in their preferred framework, or locate a dataset for training. The list has been maintained over several years and covers work spanning from roughly 2015 onward.