Semantic segmentation realization and understanding is a stringent task not just for computer vision but also in the researches of the sciences of earth, semantic segmentation decompose compound architectures in one elements, the most mutual object in a civil outside or inside senses must classified then reinforced with information meaning of all object, it’s a method for labeling and clustering point cloud automatically. Three dimensions natural scenes classification need a point cloud dataset to representation data format as input, many challenge appeared with working of 3d data like: little number, resolution and accurate of three Dimensional dataset . Deep learning now is the power and popular tool for data and image processing in computer vision, used for many applications like “image recognition”, “object detection”, “semantic segmentation”, In this research paper, provide survey a background for many techniques designed to 3 Dimensions point cloud semantic segmentation in different domains on many several available free datasets and also making a comparison between these methods.