hello, i would like to know why the inference is so slow?

[zzzzzz0407]

i modify the input size to 416*416,but the fps is just about 13?

[cory8249]

What is your CPU / GPU?

On my PC with 8700K + 1080Ti,

416 x 416
processed 40 images in 1.147 seconds
0.029 sec/image
34.88 fps

1216 x 352
processed 40 images in 1.526 seconds
0.038 sec/image
26.21 fps

BTW, kitti_detect.py is a simple demo code, maybe the overhead dominates in your case.
In my other test script, this model can reach ~90 fps @1216x352 batch=1

[zzzzzz0407]

my gpu is Titan X, and i do not modify any code, i do not know where is wrong.

[zzzzzz0407]

or if the warning is the main issue?