Xiaoqing Ye

@jhultman ,Thx, so do you think multi-scale will do no help to improve the accuracy in LIDAR-based 3D detection?

Thanks, so for those whose max disparity are less than 400, it means that I have to construct the same 400 dims , with some filling zero, am I right?...

Well , I have only a 16G memory and a single TitanX GPU and I wonder that if D is set as high as 400, will it run out of...

After carefully reading the code, I guess it comes from the 2d detection sub-network, in other words, you use the object types as a prior when feeding into the 3D...

in that case, I think the work relies much on the 2D detector 's performance. I wonder if there are ways to directly process on the point cloud. @CrazySnailer

I am also confused about the two "size_residuals" . @charlesq34

I am also curious about this, can indoor point clouds generated by RGB-D tensors adopt the spherical prediction? @BichenWuUCB

@Udonnoodle , Hi, I met a similar problem after I update the torch version to the newest one, the error is as follows: could you help me out?  My...

I solved the problem by comment 'cudnn.benchmark=True', but I don't know why.