What are the additions in OpenCV4Tegra?


I’m studying OpenCV and running examples on an Nvidia Jetson Tk1 dev board. It has the Tegra K1 SoC. Nvidia’s SDK comes with its own distr of OpenCV4Tegra. It is actually 2.4.10 with, as I understand, some additions from Nvidia.

So what’s the difference between standard OpenCV 2.4.10 and the OpenCV4Tegra?

Thank you.

In general, the “4Tegra” variants take advantage of the hardware features, e.g., without this Kepler GPU won’t contribute.

Here’s a list: http://docs.nvidia.com/gameworks/index.html#technologies/mobile/opencv_known_issues.htm%3FTocPath%3DTechnologies

Thank you for the list!

But the list introduced more questions…

  1. In the “Functions Optimized for Tegra” section does not mention the gpu module of the OpenCV.
    Does it mean that these optimized functions are actually CPU-optimized functions?

  2. The section “Known Issues / Excluded libraries” says

[b]The following table shows the features that are absent in OpenCV for Tegra.

Feature Documentation
GPU-accelerated Computer Vision http://docs.opencv.org/modules/gpu/doc/gpu.html[/b]

What exactly does this mean?? The GPU module is absent in OpenCV4Tegra? But I myself compiled and ran gpu samples with background/foreground segmentation (gpu::MOG2_GPU, etc). How should that be understood?

Section 2 says:

The GPU module, that contains CUDA-optimized functions, is available on Tegra K1 only

The Jetson TK1 Development Kit is a Tegra K1 system, and thus contains CUDA-optimized functions.