VPilot
VPilot copied to clipboard
aitorzip
Projecting points
Hi, I have successfully fetched data using DeepGTA and dataset.py. But I am still not sure how the 3D vectors FUR and BLL can be projected on to the corresponding image as shown in this video.
If you look inside https://github.com/ai-tor/DeepGTAV/blob/master/Scenario.h, there's a commented "#define DEBUG 1" line at the top. Further down you can see that, if debug is defined, it will draw the bounding boxes for the visible vehicles and pedestrians. Furthermore, there's also a switch somewhere to return the 2D instead of the 3D bounding boxes if I remember correctly.
@dominikandreas thank you for your quick reply. So in that sense it will render these boxes on the image and I cannot have them separately? And also, what's the purpose of FUR and BLL then?
There was a "bbox" flag which allowed you to do that, but apparently that has been removed. Maybe @ai-tor could comment on that
Edit: apparently there's been a bug one of the commits, so the changes that had the "bbox" flag in it got reverted. Take a look at https://github.com/ai-tor/DeepGTAV/commit/4a0cd04f8fb72fa4506a79f5b0837cae36a5010f and search for bbox to see how it was implemented before
@dominikandreas That's it, I temporary removed the commit till I have some time to look into the code and see what was originating so many crashes.
