Vassili Korotkine

It's a bit of a headache because the calibration files are different format and the paths are all different. I ended up making [a fork and refactoring the whole thing](https://github.com/vkorotkine/Pseudo-Lidar-Frustum-Raw-Inference),...

Have you tried https://stackoverflow.com/questions/34097281/convert-a-tensor-to-numpy-array-in-tensorflow ?

You need to provide it with 2D detections (e.g. computed by YOLO or detectron2). Check the `test_from_rgb_detection` scripts.

I actually had a similar issue. The approach ends up working perfectly fine, but the point cloud results are different. LIDAR (red), Your provided pseudo-lidar point cloud (green), and the...

You shouldn't have to unpack the tarball file. Just paste it in the log folder :)

Fyi, the arxiv preprint is different from the published IEEE version where this is fixed. The measurement Jacobians are also different.