I am going to do a vehicle detection using haar-like + AdaBoost in Jetson-tx1,
but couldn’t figure out the difference between OpenCV4Tegra and VisionWorks,
do these two libraries accelerate by Cuda?
which is more recommend?
thanks for replying:)
Hi Richard,
You can refer Visionworks Introduction video:
https://www.brainshark.com/nvidia/vwa-introduction?dm=5&pause=1&nrs=1
In page 17, it compares openVX(Visionworks) and OpenCV.
More, Visionworks already has a sample for feature tracking)(HARR/FAST+LK). You can give it a quick try first.
$ /usr/share/visionworks/sources/install-samples.sh ~/
$ cd ~/VisionWorks-<visworks_ver>-Samples/
$ make -j4
$ ./nvx_demo_feature_tracker
Thx a lot for the replying.
But I have more detailed questions.
OpenCV4Tegra
1.[url]http://elinux.org/Jetson/Computer_Vision_Performance[/url]
Here says
OpenCV4Tegra: A free library provided by NVIDIA containing optimizations for NVIDIA’s Tegra CPUs (ARM NEON SIMD optimizations, multi-core CPU optimizations and some GLSL GPU optimizations).
Is that means NVIDIA’s Tegra CPUs = Tegra GPU, and is accelerated by CUDA?
I run my vehicle detection on two platforms with my own cascade.xml
a.JetsonTX1—OpenCV4Tegra version 2.4.13
b.host PC-----OpenCV4Tegra version 2.4.13
How can I run vehicle detection on my host PC? I didn’t install public OpenCV, OpenCV4Tegra isn’t specially made to Tegra, but my host PC hasn’t a Tegra!
[url]https://devtalk.nvidia.com/default/topic/780052/opencv4tegra-2-4-8-whats-accelerated-/?offset=4[/url]
And host PC has more fps than Jetson TX1, the link says that imshow() is the main reason, because NVIDA doesn’t accelerate highgui in OpenCV4Tegra.
Did highgui has been accelerated in OpenCV4Tegra version2.4.13?
or other ways to show images faster?
-
No. It means one of opencv4Tegra optimization is target for ARM CPU, which will also accelerate cpu performance.
-
Could you take a look at Visionworks, which uses openvx API.
It can achieve real-time display rate.