OPRECOMP aims to build an innovative, reliable foundation for computing based on transprecision analytics. It demolishes the ultra conservative “precise” computing abstraction and replaces it with a more flexible and efficient one, namely transprecision computing.
OpenPower Summit Europe 2018 | Transprecision Computing
Dionysios Diamantopoulos, IBM Research, speaks at OpenPOWER Foundation’s OpenPOWER Summit Europe 2018.
Partners IoT Applications Videos
PULP-DroNet -- A 64mW DNN-based Visual Navigation Engine for Autonomous Nano-Drones
Fully-autonomous miniaturized robots (e.g., drones), with AI based visual navigation capabilities, are extremely challenging drivers of IoT edge intelligence capabilities. Visual navigation based on AI approaches, such as deep neural networks (DNNs) are becoming pervasive for standard-size drones. But they are considered out of reach for nano-drones with a size of a few cm2. In this work, we present the first demonstration of a navigation engine for autonomous nano-drones capable of closed-loop end-to-end DNN-based visual navigation.
Pedestrian Detection - GreenWaves Technologies
In this demonstration, GAP8 AI processor, is running Pyramidal HoG and cascade detector algorithms for pedestrian detection. GAP8 can capture and process a QVGA image (320x240px) every few minutes for several years on a small battery. The hardware required for the demo includes a GAPuino board, a QVGA Image sensor, a 3.2” LCD Screen and a 12V battery pack. Here, the camera installed outside a building can efficiently detect human presence. GAP8 can capture and process a maximum of 15fps, with its fabric controller clocked at 250 MHz and cluster at 175 MHz. In its best power per frame configuration, Gap8 delivers a power consumption of just 3.37 mW/fps.
Face Detection - GreenWaves Technologies
In this demonstration, GAP8 is running 3 Layer Pyramidal Viola-Jones Face Detection Algorithm. The hardware required for the demo includes a GAPuino board, a QVGA Image sensor, a 3.2” LCD Screen and a 12V battery pack. All enclosed in a 3D printed case. As can be seen in the video, GAP8 can reliably detect faces. GAP8 can process a maximum of 70 fps, with its fabric controller clocked at 250 MHz and cluster at 175 MHz. In its best power per frame configuration, when Gap8 detects a face it operates with a power consumption of just 0.80 mW per frame per second. If no face is present the consumption drops to 0.35mW per frame per second.