US 12,488,583 B2
Deep neural network-based object detection method, and cloud server and edge device performing deep neural network-based object detection method
Joo Chan Lee, Suwon-si (KR); and Jong Hwan Ko, Suwon-si (KR)
Assigned to Research & Business Foundation SUNGKYUNKWAN UNIVERSITY, Suwon-si (KR)
Filed by Research & Business Foundation SUNGKYUNKWAN UNIVERSITY, Suwon-si (KR)
Filed on May 16, 2023, as Appl. No. 18/197,891.
Claims priority of application No. 10-2022-0059423 (KR), filed on May 16, 2022.
Prior Publication US 2023/0368518 A1, Nov. 16, 2023
Int. Cl. G06V 10/94 (2022.01); G06V 10/77 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01)
CPC G06V 10/94 (2022.01) [G06V 10/7715 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01)] 15 Claims
OG exemplary drawing
 
1. A cloud server for detecting an object of an image based on a single feature received from an edge device, comprising:
a memory configured to store information on a pre-trained extractor of the cloud server having a first layer and a second layer and a pre-trained reconstructor of the cloud server; and
a processor,
wherein the processor is configured to:
receive a single feature of an image from the edge device, wherein the edge device includes an extractor having a plurality of layers and the single feature of the image is extracted by a first layer of the extractor of the edge device corresponding to a predetermined split point among the plurality of layers of the extractor of the edge device; and
obtain an output feature of the extractor of the cloud server by inputting the single feature received from the first layer included in the extractor of the edge device to the second layer included in the extractor of the cloud server, and
detect an object in the image by inputting the output feature obtained from the extractor of the cloud server to the reconstructor of the cloud server, wherein a number of channels of at least one of the plurality of layers included in the extractor of the edge device or the extractor of the cloud server is less than a number of channels corresponding to at least one of a plurality of layers included in the reconstructor of the cloud server.