Artificial intelligence-based techniques for automated visual data searching using edge devices

Patent metadata

Field	Value
Publication number	US-12481695-B2
Application number	US-202217718803-A
Country	US
Kind code	B2
Filing date	Apr 12, 2022
Priority date	Apr 12, 2022
Publication date	Nov 25, 2025
Grant date	Nov 25, 2025

Abstract

Methods, apparatus, and processor-readable storage media for artificial intelligence-based techniques for automated visual data searching using edge devices are provided herein. An example computer-implemented method includes obtaining visual data from one or more edge devices; generating at least one automated searching tool by processing at least a portion of the obtained data using one or more artificial intelligence techniques; deploying the at least one automated searching tool to at least a portion of the one or more edge devices; and performing one or more automated actions based at least in part on data received, from at least a portion of the one or more edge devices, in connection with operation of the at least one automated searching tool.

First claim

What is claimed is:

1 . A computer-implemented method comprising:
obtaining visual data from one or more edge devices; generating at least one automated searching tool by processing at least a portion of the obtained visual data using one or more artificial intelligence techniques, wherein generating the at least one automated searching tool comprises:
extracting content from the obtained visual data as one or more embedding vectors using at least one deep learning embedding technique; and
defining multiple portions of the one or more embedding vectors for use in matching with one or more portions of input images, wherein the multiple portions of the one or more embedding vectors are defined as a plurality of distinct data structures representing distinct content associated with the one or more embedding vectors, and wherein the plurality of distinct data structures each comprise at least one distance metric, at least one centroid calculation result, and at least one distance threshold value;
deploying the at least one automated searching tool to at least a portion of the one or more edge devices; and performing one or more automated actions based at least in part on data received, from the at least a portion of the one or more edge devices, in connection with operation of the at least one automated searching tool; wherein the method is performed by at least one processing device comprising a processor coupled to a memory.

2 . The computer-implemented method of claim 1 , wherein extracting the content from the obtained visual data as the one or more embedding vectors using the at least one deep learning embedding technique comprises processing the at least a portion of the obtained visual data through multiple feature extraction-related layers of a deep learning network, wherein the multiple feature extraction-related layers comprise one or more convolution and rectified linear unit activation function layers, one or more pooling layers, one or more flatten layers, and one or more fully connected layers.

3 . The computer-implemented method of claim 2 , wherein extracting the content from the obtained visual data as the one or more embedding vectors using the at least one deep learning embedding technique further comprises processing output from the multiple feature extraction-related layers using one or more output layers, wherein the one or more output layers comprise one or more of at least one nearest neighbor algorithm, at least one softmax function, and at least one regression function.

4 . The computer-implemented method of claim 1 , wherein generating the at least one automated searching tool further comprises defining one or more similarity thresholds for the one or more embedding vectors.

5 . The computer-implemented method of claim 1 , wherein performing the one or more automated actions comprises determining whether at least a portion of the data received from the at least a portion of the one or more edge devices corresponds to at least one of multiple areas.

6 . The computer-implemented method of claim 1 , wherein the at least one automated searching tool comprises a given number of float numbers that is below a predetermined threshold.

7 . The computer-implemented method of claim 1 , wherein performing the one or more automated actions comprises automatically generating and outputting at least one notification to at least one external system, wherein the at least one notification pertains at least to geographic information associated with the data received from the at least a portion of the one or more edge devices in connection with the operation of the at least one automated searching tool.

8 . The computer-implemented method of claim 1 , wherein performing the one or more automated actions comprises automatically training the one or more artificial intelligence techniques using at least a portion of the data received from the at least a portion of the one or more edge devices in connection with the operation of the at least one automated searching tool.

9 . The computer-implemented method of claim 1 , wherein obtaining the visual data from the one or more edge devices comprises obtaining at least one of image data, video data, and text data from the one or more edge devices.

10 . A non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing device causes the at least one processing device:
to obtain visual data from one or more edge devices; to generate at least one automated searching tool by processing at least a portion of the obtained visual data using one or more artificial intelligence techniques, wherein generating the at least one automated searching tool comprises:
extracting content from the obtained visual data as one or more embedding vectors using at least one deep learning embedding technique; and
defining multiple portions of the one or more embedding vectors for use in matching with one or more portions of input images, wherein the multiple portions of the one or more embedding vectors are defined as a plurality of distinct data structures representing distinct content associated with the one or more embedding vectors, and wherein the plurality of distinct data structures each comprise at least one distance metric, at least one centroid calculation result, and at least one distance threshold value;
to deploy the at least one automated searching tool to at least a portion of the one or more edge devices; and to perform one or more automated actions based at least in part on data received, from the at least a portion of the one or more edge devices, in connection with operation of the at least one automated searching tool.

11 . The non-transitory processor-readable storage medium of claim 10 , wherein extracting the content from the obtained visual data as the one or more embedding vectors using the at least one deep learning embedding technique comprises processing the at least a portion of the obtained visual data through multiple feature extraction-related layers of a deep learning network, wherein the multiple feature extraction-related layers comprise one or more convolution and rectified linear unit activation function layers, one or more pooling layers, one or more flatten layers, and one or more fully connected layers.

12 . The non-transitory processor-readable storage medium of claim 11 , wherein extracting the content from the obtained visual data as the one or more embedding vectors using the at least one deep learning embedding technique further comprises processing output from the multiple feature extraction-related layers using one or more output layers, wherein the one or more output layers comprise one or more of at least one nearest neighbor algorithm, at least one softmax function, and at least one regression function.

13 . The non-transitory processor-readable storage medium of claim 10 , wherein performing the one or more automated actions comprises automatically training the one or more artificial intelligence techniques using at least a portion of the data received from the at least a portion of the one or more edge devices in connection with the operation of the at least one automated searching tool.

14 . An apparatus comprising:
at least one processing device comprising a processor coupled to a memory; the at least one processing device being configured:
to obtain visual data from one or more edge devices;
to generate at least one automated searching tool by processing at least a portion of the obtained visual data using one or mo

Assignees

• Dell Products Lp

Inventors

• Gong Min
• Bao Qi
• Qiu Qicheng
• Wu Daniel
• Huang Shuangchen
• Wang Rui

Classifications

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06F18/24137

  Distances to cluster centroïds · CPC title

• G06V10/454

  Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN] · CPC title

• G06F17/18 Primary

  for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} · CPC title

• G06V10/82

  using neural networks · CPC title