TX2: caffe model that runs slower on nvcaffe GPU than on OpenCV CPU

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

This caffe model:
https://github.com/spmallick/learnopencv/blob/master/FaceDetectionComparison/models/res10_300x300_ssd_iter_140000_fp16.caffemodel

runs slower on nvcaffe GPU than on OpenCV CPU

prototxt:
https://github.com/spmallick/learnopencv/blob/master/FaceDetectionComparison/models/deploy.prototxt

OpenCV code here:
https://github.com/spmallick/learnopencv/blob/master/FaceDetectionComparison/face_detection_opencv_dnn.cpp

2

Hi,

Have you maximized the CPU/GPU clocks?

sudo jetson_clocks.sh

To get better performance, it’s recommended to use TensorRT:
https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html

Thanks.

3

Many thanks! First time I see the fan blowing.