Latest with the release of the iPhone X bringing deep learning and its application to solve computer vision problems to mobile (edge) devices, object detection and the state-of-the-art in-face detection-accuracy took an enormous leap forward.
The challenge we face now is that the learned models ported to the edge device require an order of magnitudes more memory, more disk storage and more computational resources.
To address those resource problems, we unveil the first prototype of NEMA®|xNN, a power efficient inference-accelerator, to solve computer-vision tasks in edge-computing applications for convolutional network (CNN) tasks.
Developing power-optimized graphics software is inherently complex; therefore, smart analysis and programming
tools become crucial instruments for the programmer.
Think Silicon® is adding two new members to its SDK, making programmers’ life easier. NEMA®|Power-Model enables programmers to identify and optimize the most energy consuming parts in their graphics software application. NEMA®|SHADER-Edit is a vertex and fragment shading editor with an integrated compiler, allowing programmers to rapidly create and compile optimal GLSL shaders offline without the need for a deeper knowledge of the internal 3D GPU operations.
Come and visit us at “Embedded World” in Nuremberg/Germany (Feb 27 - Mar 1st, Hall 4 Booth 4-673) so that you can see NEMA®|xNN, NEMA®|Power-Model and NEMA®|SHADER-Edit live on our booth.
Think Silicon has answers!
Ulli Mueller, VP Marketing & Business Development, Think Silicon