Hailo Demonstration of Highly Power-efficient Full HD Segmentation Using the Hailo-8 DL Processor
2.1 هزار بار بازدید -
5 سال پیش
-
For more information about embedded
For more information about embedded vision, including hundreds of additional videos, please visit http://www.embedded-vision.com.
Liran Bar, Director of Product Marketing at Hailo, demonstrates the company's latest embedded vision technologies and products at the 2019 Embedded Vision Summit. Specifically, Bar demonstrates full HD semantic segmentation with high power efficiency using the Hailo-8 DL processor for edge devices.
The Hailo-8 DL processor for edge devices features up to 26 TOPS and significantly outperforms all other edge processors, with area and power efficiency far superior to other leading solutions by a considerable order of magnitude, and all at a size smaller than a penny (including the required memory).
In this semantic segmentation demo, a stream of FHD (1920 x 1080) video is injected from a PC via a 1 GbE interface to the Hailo-8 for segmentation. The Hailo-8 is running both the entire ResNet-18 and FCN-16 networks (without pruning in both cases) at 21 FPS while consuming only 1.2 W, and with power efficiency of 2.6 TOPS/W. The performance in this demo is limited by the 1 GbE interface.
For more information please visit https://www.hailo.ai
Liran Bar, Director of Product Marketing at Hailo, demonstrates the company's latest embedded vision technologies and products at the 2019 Embedded Vision Summit. Specifically, Bar demonstrates full HD semantic segmentation with high power efficiency using the Hailo-8 DL processor for edge devices.
The Hailo-8 DL processor for edge devices features up to 26 TOPS and significantly outperforms all other edge processors, with area and power efficiency far superior to other leading solutions by a considerable order of magnitude, and all at a size smaller than a penny (including the required memory).
In this semantic segmentation demo, a stream of FHD (1920 x 1080) video is injected from a PC via a 1 GbE interface to the Hailo-8 for segmentation. The Hailo-8 is running both the entire ResNet-18 and FCN-16 networks (without pruning in both cases) at 21 FPS while consuming only 1.2 W, and with power efficiency of 2.6 TOPS/W. The performance in this demo is limited by the 1 GbE interface.
For more information please visit https://www.hailo.ai
5 سال پیش
در تاریخ 1398/04/03 منتشر شده
است.
2,150
بـار بازدید شده