search

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 330 forks source link

TensorRT model consumes higher memory than the MXNet model #67

Open nsabir2011 opened 4 years ago

nsabir2011 commented 4 years ago

After Successfully converting the TensorRT model and running it, the resource consumption seems to be higher. How do i optimize it to be lower than the MXNet model?

TensorRT VRAM Usage: 1019 MiB TensorRT RAM Usage: 2.0 GiB

MXNet VRAM Usage: 745 Mib MXNet RAM Usage: 1.7 GiB

I have Cuda 10.0, cuDNN 7.5.1, TensorRT 6.0.1.5 installed.