search

charlesq34 / pointnet

PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation
Other
4.66k stars 1.44k forks source link

Room Dimensions coming out incorrect #214

Open Naor92 opened 4 years ago

Naor92 commented 4 years ago

Hi everyone, I'm trying to run segmentation on my own .txt files with the following format XYZRGB.

The segmentation does run but it seems like the .obj files being spit out have incorrect bounding box size.

Screenshot from 2019-10-30 13-51-11

You can see in the screenshot that the bounding box leaves half of the point cloud outside of the calculated area. I have made the following adjustments to the batch_inference code:

Get room dimension..

data_label = np.loadtxt(room_path)
data = data_label[:,0:6]
max_room_x = max(data[:,0], key=abs)
max_room_y = max(data[:,1], key=abs)
max_room_z = max(data[:,2], key=abs)

But that only seems to stretch the room dimension and points inside. How can I alter the code to ensure all points in the cloud are segmentated? Appreciate any insight anyone can provide.