DoubangoTelecom / ultimateALPR-SDK

World's fastest ANPR / ALPR implementation for CPUs, GPUs, VPUs and NPUs using deep learning (Tensorflow, Tensorflow lite, TensorRT, OpenVX, OpenVINO). Multi-Charset (Latin, Korean, Chinese) & Multi-OS (Jetson, Android, Raspberry Pi, Linux, Windows) & Multi-Arch (ARM, x86).
https://www.doubango.org/webapps/alpr/
Other
617 stars 158 forks source link

Benchmark performance #155

Open Nook2007 opened 3 years ago

Nook2007 commented 3 years ago

c++ 2x faster than java on linux x64, with same machine/rate/config/images

DoubangoTelecom commented 3 years ago

X86-64 or ARM64? Both (Java and C++) are using parallel (or sequential) mode? Please share sample code you're using.

Nook2007 commented 3 years ago

recognizer.zip Tested on x86-64 sequential mode

DoubangoTelecom commented 3 years ago

Please attach both logs (c++ and java)

Nook2007 commented 3 years ago

java.txt c++.txt

Nook2007 commented 3 years ago

c++.txt java.txt updated logs with same config

DoubangoTelecom commented 3 years ago

The only difference I see is the input: C++ uses RGB while Java uses BGR. In your Java code change ULTALPR_SDK_IMAGE_TYPE_BGRA32 to ULTALPR_SDK_IMAGE_TYPE_RGBA32 and ULTALPR_SDK_IMAGE_TYPE_BGR24 to ULTALPR_SDK_IMAGE_TYPE_RGB24 and share the Java logs again.

The change I'm asking to do isn't correct in production, so revert the change after collecting the logs.

Nook2007 commented 3 years ago

java.txt nothing changed for rgb24

Nook2007 commented 2 years ago

Any updates?