id	author	title	date	pages	extension	mime	words	sentence	flesch	summary	cache	txt
fcis-25709	Wang, Zishuo; Lai , Tianxiang; Wang, Yufeng; Gao, Xingquan	3D Point Cloud Semantic Segmentation based on Multi-scale Dense Nested Networks	2024	9	.pdf	application/pdf	6126	308	55	Compared with methods such as farthest point sampling[19], and inverse density sampling [20], MDNN uses random sampling operations to reduce the spatial resolution of point features, thereby improving computational efficiency and abstraction. Then, along the direction of the encoder path, the network gradually sparse the point cloud (where N is the number of points) in the order of (N→N/4→N/16→N/64→N/256→N/512), each level of MFFM learns the multi-scale of point features at multiple network levels, and gradually expands the receiving domain to obtain higher level of abstract semantic features.	cache/fcis-25709.pdf	txt/fcis-25709.txt