I am using the sample OpenCV code for tracking facial features but am only getting 4 FPS using a front facing camera. It would be great to get higher performance in the 20-30 FPS range which I think is possible.
My progress so far is on my website.
https://sites.google.com/site/biothemonitors/home/android-programming-labs
I don't know if you already found the good answer or not, but the best thing to do is reduce the resolution to about 320x240 for the camera. This will make your window quite small, but you can "stretch" it back to the full screen of your phone by making the matrix into a bitmap and scaling it, then convert it back to a matrix. This did the trick for my Android phone, your FPS will be a lot better.
This is a good idea and I will explore it. The only drawback of using this approach could be a lack of resolution for the facial detection since I eventually want to be able to detect the eyes for clicking of the mouse.
that is true, I found that if you use Tutorial 2 (Mixed process) from the OpenCV examples that it doubles your FPS. I have no idea how this comes, but my app went from 3 FPs to about 8 FPS when using this tutorial. I think the use of the JNI part in this example makes the phone give it more memory to process everything (not sure about this though). I haven't found a way to get better FPS, but maybe this link will help you http://answers.opencv.org/question/9105/best-way-to-get-most-fps-android-camera/
Also you can take a look at:
https://developer.qualcomm.com/computer-vision-fastcv
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training