GPU Batched NMS
This commit is contained in:
@@ -75,7 +75,7 @@ cudaError_t sortDetections(
|
||||
int threads_per_block = 0;
|
||||
int number_of_blocks = 4;
|
||||
|
||||
if (_topK % 2 == 0 || _topK >= number_of_blocks)
|
||||
if (_topK % 2 == 0 && _topK >= number_of_blocks)
|
||||
threads_per_block = _topK / number_of_blocks;
|
||||
else
|
||||
threads_per_block = (_topK / number_of_blocks) + 1;
|
||||
|
||||
Reference in New Issue
Block a user