Luc Vedrenne
Luc Vedrenne
Just to add a little clarification to the answer just above: PointNet is exactly invariant to point duplication in the absence of Batch Normalization layer. With BatchNorm, this isn't true...
For anyone interested in this feature, a basic workaround is to retrieve the bounding box coordinates afterward using template matching: assume you have your original image in an `image` variable...
What's your CUDA version ? If it is < 11, you should consider upgrade it, try and see if it fixes the error.
Hi ! You can use the groundtruth poses in the 'gt.log' file to align a pair of point clouds.