I’m using generator.py to explore generating simulated data. An example row in the outputted kitti label folder is:
palette 0.00 3 0.00 73 373 140 396 0.00 0.00 0.00 0.00 0.00 0.00 0.00
(The ‘3’ means ‘unknown’ occusion, the four floats are the bounding box in pixels. Link to definition)
I’m interested in obtaining the ‘location’ of the object in camera coordinates. I can then use this for training object pose detection.
Is there any way to get this information outputted? Any ideas?
Okay, a little deeper. This won’t help. Seems Pose is a two step process - first object detection, then pose inside that cropped bounding box (trained from a CAD model of your object.)
Unfortunately the last version of the Object Post Estimator tutorial isn’t available/doesn’t work. (For unity3D.) Pieces seem to exist in Isaac SDK.
However, it doesn’t seem that Kitti data will be useful for the pose element. Sorry to bother.