Performance
latency without preprocess: 0.82ms
latency with preprocess: 1.26ms
top5 error: 0.0699
Enviroment Prepare
Host: 1 nvidia Tesla T4 , 1 Xeon(G) Gold 6130 cpu , 128G cpu memory
Software: ubuntu 16.04 , nvidia driver 410.104 , cuda 10.0 , cudnn 7.5 , tensorrt 5.1.5.0
Model: caffe model resnet50
Usage
Modify 'model_path, imgs_path, gt_path' in test.cpp.
mkdir build && cd build && cmake ..
make
./AIProject