US 11,741,568 B2
Systems and methods for low-power, real-time object detection
Haofeng Kou, Palo Alto, CA (US); Kuipeng Wang, Beijing (CN); Le Kang, Mountain View, CA (US); Xuejun Wang, Guangdong (CN); and Yingze Bao, Mountain View, CA (US)
Assigned to Baidu USA LLC, Sunnyvale, CA (US); and Baidu.com Times Technology (Beijing) Co., Ltd., Beijing (CN)
Appl. No. 16/770,063
Filed by Baidu USA, LLC, Sunnyvale, CA (US); and Baidu.com Times Technology (Beijing) Co., Ltd., Beijing (CN)
PCT Filed Jun. 29, 2018, PCT No. PCT/CN2018/093717
§ 371(c)(1), (2) Date Jun. 4, 2020,
PCT Pub. No. WO2020/000383, PCT Pub. Date Jan. 2, 2020.
Prior Publication US 2021/0173707 A1, Jun. 10, 2021
Int. Cl. G06F 9/48 (2006.01); G06T 1/20 (2006.01); G06F 1/329 (2019.01); G06F 9/50 (2006.01); G06N 3/08 (2023.01); G06F 18/214 (2023.01)
CPC G06T 1/20 (2013.01) [G06F 1/329 (2013.01); G06F 9/4893 (2013.01); G06F 9/5038 (2013.01); G06F 18/214 (2023.01); G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. An apparatus comprising:
a camera sensor to capture one or more image frames;
a memory storing at least a first object-detection model and a second object-detection model; and
a processor comprising a first CPU core, a second CPU core, and a plurality of shaves for processing, in which:
the first CPU core is configured to run a first instance of an operating system to handle general CPU tasks for the apparatus; and
the second CPU core is configured to run a second instance of an operating system to handle the image frames and to schedule computation tasks to be implemented by the plurality of shaves for the first object-detection model and the second object-detection model to detect an object in one or more image frames.