AlexNet and Google net accuracy problem

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Hi,

when I try to validate AlexNet and GoogleNet accuracy from tensorRT I get worse results compared to tensorflow (~5-10%). I use the same preprocessing and official bvlc models (imageNet).

Also when I try to use int8 calibration with 1000 images from validation set I get 0 accuracy on both networks.

Thanks.