US 12,130,859 B2
Method and system for feature based image retrieval
Preetam Prabhu Srikar Dammu, Hyderabad (IN); Srinivasa Rao Chalamala, Hyderabad (IN); and Ajeet Kumar Singh, Pune (IN)
Assigned to TATA CONSULTANCY SERVICES LIMITED, Mumbai (IN)
Filed by Tata Consultancy Services Limited, Mumbai (IN)
Filed on Apr. 21, 2022, as Appl. No. 17/660,034.
Claims priority of application No. 202121021834 (IN), filed on May 14, 2021.
Prior Publication US 2022/0365963 A1, Nov. 17, 2022
Int. Cl. G06F 16/583 (2019.01); G06F 16/56 (2019.01); G06V 10/25 (2022.01)
CPC G06F 16/583 (2019.01) [G06F 16/56 (2019.01); G06V 10/25 (2022.01)] 12 Claims
OG exemplary drawing
 
1. A processor implemented method comprising:
receiving, by one or more hardware processors, an input image and a corresponding query label, wherein the query label indicates a Region of Interest (ROI);
selecting, by the one or more hardware processors, a feature specific encoder from a plurality of feature specific encoders based on the ROI, wherein the plurality of feature specific encoders are pretrained using a mask based loss function;
computing, by the one or more hardware processors, a first set of feature vectors corresponding to the ROI based on the selected feature specific encoder;
computing, by the one or more hardware processors, a feature specific Locality Sensitive Hashing (LSH) value corresponding to the ROI based on the first set of feature vectors; and
obtaining, by the one or more hardware processors, a plurality of matching images from a plurality of database images based on a comparison between the computed feature specific LSH value and a plurality of feature specific LSH values stored in a feature specific LSH database, wherein the feature specific LSH database is generated based on the plurality of database images and the plurality of feature specific encoders.