Point cloud picture is not the common point cloud form, why we can use point transformer to process it?

nomewang / M3DM

MIT License

140 stars 18 forks source link

Point cloud picture is not the common point cloud form, why we can use point transformer to process it? #27

Closed limaodaxia closed 8 months ago

limaodaxia commented 10 months ago

Usually, point clouds form are like modelnet40 in which one point is described as an array containing [x, y, z, ...]. But in MVTec 3D AD dataset, the point cloud is a one channel image. What is the relationship between the two?