US 12,437,507 B2
Image analysis method, image analysis apparatus, and image analysis system
Xiaoning Wang, Shenzhen (CN); Jianhua Yao, Shenzhen (CN); Hong Shang, Shenzhen (CN); and Han Zheng, Shenzhen (CN)
Assigned to TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed by TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED, Shenzhen (CN)
Filed on Mar. 18, 2022, as Appl. No. 17/699,056.
Application 17/699,056 is a continuation of application No. PCT/CN2021/080185, filed on Mar. 11, 2021.
Claims priority of application No. 202010250222.4 (CN), filed on Apr. 1, 2020.
Prior Publication US 2022/0207862 A1, Jun. 30, 2022
Int. Cl. G06V 10/00 (2022.01); G06V 10/22 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01)
CPC G06V 10/764 (2022.01) [G06V 10/225 (2022.01); G06V 10/774 (2022.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01); G06V 2201/03 (2022.01); G06V 2201/07 (2022.01)] 16 Claims
OG exemplary drawing
 
1. An image analysis method, performed by an image analysis apparatus and comprising:
obtaining an image, the image being an endoscopic tissue image;
performing image classification on the image by using an image classification network, to obtain an image classification result of an image category of the image, the image category including a first category and a second category different from the first category, the first category being a lesion category indicating that the image includes at least one lesion region, and the second category being a non-lesion category indicating that the image does not include a lesion region;
performing object detection on the image by using an object detection network, to obtain an object detection result of a target object associated with the first category, the object detection result indicating whether a candidate lesion region corresponding to the lesion category is detected by the object detection network; and
generating an image analysis result of the image based on the image classification result and the object detection result, comprising:
in response to the image classification result indicating that the image category is the non-lesion category and the object detection result indicating that the candidate lesion region is detected, comparing a first confidence level corresponding to the second category with a first threshold, and indicating, by the image analysis result, whether the image has the target object based on whether the first confidence level is less than or equal to the first threshold; and
in response to the image classification result indicating that the image category is the lesion category and the object detection result indicating that the candidate lesion region is not detected, comparing a second confidence level corresponding to the first category with a second threshold, and indicating, by the image analysis result, whether the image has the target object based on whether the second confidence level is greater than the second threshold.