search

charlesq34 / frustum-pointnets

Frustum PointNets for 3D Object Detection from RGB-D Data
Apache License 2.0
1.59k stars 538 forks source link

How can I use the model if the point cloud data lacking of intensity #73

Open xhuan28 opened 5 years ago

xhuan28 commented 5 years ago

The point cloud data for both training and testing in this project is expected to have 4 channels, X, Y, Z and intensity. However, the point could data on my hand is sourced from RGBD camera, which is lacking of intensity value. How can I use the model with only X,Y,Z and RGB information?

lambders commented 5 years ago

I think you'll have to retrain the model -- there's an argument '--no-intensity' in the training script which allows you to train only using the XYZ channels of the pointcloud: https://github.com/charlesq34/frustum-pointnets/blob/2ffdd345e1fce4775ecb508d207e0ad465bcca80/train/train.py#L34

xhuan28 commented 5 years ago

@amanda321 Thanks for your advice. I didn't notice this argument. Now I am retraining the model with '--no_intensity' option.