What is OpenCV

OpenCV is an open source computer vision library intended for image processing.
It is provided free of charge for both educational and commercial purposes under the BSD (Berkeley Software Distribution) license.

In recent years, it has become common to see applied products that support a safe and comfortable life by using cameras as human eyes. Have you ever seen in the news about products that apply the following image processing technologies?

Devices such as these applied products require advanced image processing technology, which is evolving day by day.

Also, in order to "detect an object" from the image data information output from the camera, knowledge of a complicated mathematical algorithm is required, but if you write a few lines of program in OpenCV, you can easily detect the object. can be realized.

What you can do with OpenCV

If you want to know more about OpenCV, please click here.
OpenCV: Introduction

OpenCV performance

The following two graphs are published on TI's website.
For Arm cores and DSP cores on a particular device, you can read from the graph how much the OpenCV routines are sped up when code is moved from Arm cores to DSP cores.

Note that the performance of the Cortex-A15 core is standardized at 1, and the performance of the C66x core is shown relative to the Cortex-A15. Also, this comparison takes into account the number of cores and the speed of the processor.

When benchmark cores are 2x Arm A15 (1.5GHz) and 2x C66x DSP (750MHz)

When benchmark cores are 4x Arm A15 (1.2GHz) and 8x C66x DSP (1.2GHz)

* From the user's point of view, the acceleration using the DSP core is transparent. Measured performance is the result of the compiler at a point in time, not optimized for DSP.

TI also has a table showing the relative performance improvement gained by offloading from the device's Arm core to the DSP core, which you can read more about here.
OpenCV library (TI official site)

TI processor compatible with OpenCV

TI's products fully support OpenCV 3.1 on all Arm® Cortex®-A based processors.

This includes the AM335x, AM437x, and AM57xx Sitara processor families, as well as the 66AK2x multicore DSP + Arm processor family of processors. Also, all OpenCV features support multi-core Arm for processors with multiple Arm Cortex-A15 cores.

At the end

What did you think. Using TI's Sitara processor family introduced this time, I actually tried pedestrian detection. Since it introduces everything from environment preparation to operation results, please take a look at this article as well.

Let's detect people with OpenCV compatible Arm (R) processor

Contact Us