[introduction to point cloud dataset]

Classification on ModelNet40

ModelNet40 [55] was a commonly used object classification dataset for 3D computer graphics CAD models. It has 40 object categories, each of which contains 100 unique CAD models

3D Object Classification on ScanObjectNN

3D Object Classification on ScanObjectNN ScanObjectNN [49] contains about 15, 000 real scanned objects that are categorized into 15 classes with 2, 902 unique object instances. Due to occlusions and noise, ScanObjectNN poses significant challenges to existing point cloud analysis methods

3D Object Part Segmentation on ShapeNetPart

ShapeNetPart [59] is an object-level dataset for part segmentation. It consists of 16, 880 models from 16 different shape categories, 2-6 parts for each category, and 50 part labels in total

3D Semantic Segmentation on S3DIS

S3DIS [1] (Stanford Large-Scale 3D Indoor Spaces) is a challenging benchmark composed of 6 large-scale indoor areas, 271 rooms, and 13 semantic categories in total