No performance gain with TensorRT 5

YonghaoHe / LFFD-A-Light-and-Fast-Face-Detector-for-Edge-Devices

A light and fast one class detection framework for edge devices. We provide face detector, head detector, pedestrian detector, vehicle detector......

MIT License

1.31k stars 329 forks source link

No performance gain with TensorRT 5 #90

Open pedromoraesh opened 4 years ago

pedromoraesh commented 4 years ago

Hey Guys!

I did run the Mxnet to ONNX script and then running predictions with TensorRT.

My work deals with video processing so i couting the processing time for each frame. Without ONNX, running pure MXNET, got 7~9ms per frame. Predicting with TensorRT i got same result!

Running in a LAPTOP with GTX1060, TensorRT 5.1.5 and CUDA 10.1 (same as reported in this repo)

Did someone experience something like that?