Gabriele's profile - activity

Thanks Alexander!

Thanks for your reply Alexander. On the one hand, with the Android version I could enable Tegra optimization. On the other hand, I'm wondering whether the Android version is slower than the Linux one, for example because of the Java JNI layer. In a nutshell, I'm trying to understand whether it's worth replacing my Linux-based installation with an Android one just to enable Tegra optimization. Is there any documented comparison about the two versions?

PowerVR does support OpenCL 1.0, still it is up to the main system-on-chip producer to provide drivers enabling its support. As an example, even if Texas Instruments embeds this GPU within its OMAP4 architecture, no driver is available at the moment, which in turns implies that no OpenCL support is available

Hi,

I noticed an impressive demo about OpenCV for Tegra (http://www.embedded-vision.com/industry-analysis/video-interviews-demos/2012/12/19/december-2012-embedded-vision-alliance-member--0?goback=%2Egde_3934172_member_197847664). However, this version only runs on Android, and not on Linux distros. Assuming that I own a processing board with ARM CPU and Tegra GPU, what OpenCV version would you suggest me to use in terms of processing performance: (i) OpenCV on Linux or (ii) OpenCV for Tegra on Android? Have you ever compared the two different versions?

Thanks Gabriele

Hi,

I'm adopting SimpleBlobDetector to detect a set of blobs, by calling:

vector<keypoint> keypoints; blobDetector->detect(frame,keypoints);

However, I can't find any method to retrieve and draw the contours associated to each blob. I can only draw their centroids, by using:

drawKeypoints(input, keypoints, output, CV_RGB(255,0,0),cv::DrawMatchesFlags::DEFAULT);

Digging into the SimpleBlobDetector source code, it seems to me that findContours is called, by the result is stored in a local variable, thereby not available with any public method. Am I right?

Thanks Gabriele