Parallel_NMS
Parallel_NMS copied to clipboard
large number of bounding boxes
Since the maximum number of cuda threads is 1024.
So if the number of bounding boxes exceeds 1024 (count > 1024), the current implementation won't work.
https://github.com/jeetkanjani7/Parallel_NMS/blob/master/GPU/nms.cu#L104
Any idea?