This demo highlights the benefits of using GPU Compute via OpenCL™ to enhance detection rates and therefore precision of gesture recognition. By using OpenCL the detection rate can be increased from ~10 detections per second with a purely CPU-based algorithm to ~40 detections per second when OpenCL is used.
The demo runs on the Arndale development board which contains the Samsung Exynos 5 Dual SoC. This SoC contains a dual core Arm Cortex-A15 CPU and quad core Arm Mali-T604 GPU.
Notice how the gesture detection algorithm is still able to function when the camera feed is replaced with a very low quality video shot in low light conditions. This low quality video simulates what the standard webcam sees and also has extra noise added to make the test more challenging.
eyeSight's Gesture Recognition & Mali: Touch-Free Capabilities Optimized
Read eyeSight's blog on Arm's Connected Community to learn more about this demo
Smile to the camera, it's OpenCL!
Learn about the Huawei Honor 7 the first smartphone to optimize all it's photos with OpenCL on Mali.
Realizing the benefits of GPU Compute for real application with Mali GPUs
An blog on the wide arrange of real use cases being addressed by OpenCL with Mali.
Seeing the future with Computer Vision
How machines are interpreting the outside world; including Arm's acquisition of Apical.
eyeSight Technology's website
See the latest from eyeSight on their website.