US 12,430,902 B2
Apparatus and method for detecting elements of an assembly
Orlando Grigoriadis, Newbury (GB); and Nazmi Shafin, Newbury (GB)
Assigned to Vodafone Group Services Limited, Newbury (GB)
Appl. No. 17/999,836
Filed by Vodafone Group Services Limited, Newbury (GB)
PCT Filed May 24, 2021, PCT No. PCT/GB2021/051259
§ 371(c)(1), (2) Date Nov. 23, 2022,
PCT Pub. No. WO2021/240137, PCT Pub. Date Dec. 2, 2021.
Claims priority of application No. 20386029 (EP), filed on May 29, 2020.
Prior Publication US 2023/0215154 A1, Jul. 6, 2023
Int. Cl. G06V 10/00 (2022.01); G06V 10/80 (2022.01); G06V 10/82 (2022.01); G06V 30/10 (2022.01); G06V 30/18 (2022.01)
CPC G06V 10/82 (2022.01) [G06V 10/809 (2022.01); G06V 30/10 (2022.01); G06V 30/18057 (2022.01)] 17 Claims
OG exemplary drawing
 
1. An apparatus for detecting elements of an assembly, comprising:
one or more processors; and
memory storing first and second artificially intelligent classifiers for detecting elements in a high-resolution image of the assembly, wherein the first artificially intelligent classifier is pre-trained to detect first elements and the second artificially intelligent classifier is pre-trained to detect second elements, each of the first elements having a size within a first size range, and each of the second elements having a size within a second size range, in which the first size range includes elements having a size that is greater than the size of elements included within the second size range;
the memory comprising instructions which, when executed by one or more of the processors, cause the processor(s) to:
receive a high resolution image of the assembly;
sequentially use the first classifier and then the second artificially intelligent classifier to:
create plural scaled sub-images taken from the high resolution image, each sub-image having a size of at least the maximum size of the respective first or second elements, and scaled to be resized to a pixel input image size of the artificially intelligent classifier;
ingest the scaled sub-images into the artificially intelligent classifier to search for elements in the scaled sub-images; and
obtain, from the output of the artificially intelligent classifier, data indicative of bounding box coordinates in the high resolution image corresponding to detected first and then second elements;
the memory further comprising instructions to cause the processor(s) to prevent the second artificially intelligent classifier from subsequently searching within the bounding boxes previously obtained by the first artificially intelligent classifier.