Image fingerprinting conversion between different image fingerprinting models

A system, comprising:

• one or more processors;
• and
• one or more computer-readable memory devices, comprising:
• a database storing a plurality of images,
• a first image fingerprinting model trained to generate a first image fingerprint embedding comprising a multi-dimensional feature vector representing visual aspects of an input image,
• a second image fingerprinting model trained to generate a second image fingerprint embedding comprising a multi-dimensional feature vector representing visual aspects of an input image,
• an untrained converter model,
• a converter training component comprising instructions that, upon execution by the one or more processors, cause the one or more processors to:
• generate a plurality of first image fingerprint embeddings, the instructions to generate comprising instructions that cause the one or more processors to:
• providing each of a plurality of images as input to the first image fingerprinting model, and
• receive each of the plurality of first image fingerprint embeddings as output from the first image fingerprinting model;
• generate a plurality of second image fingerprint embeddings, the instructions to generate comprising instructions that cause the one or more processors to:
• provide each of the plurality of images as input to the second image fingerprinting model, and
• receive each of the plurality of second image fingerprint embeddings as output from the second image fingerprinting model, wherein the first image fingerprint embedding is different than the second image fingerprint embedding generated for each of the plurality of images;
• and
• for each image of the plurality of images:
• provide the first image fingerprint embedding associated with the respective image as input to the untrained converter model,
• receive an output image fingerprint embedding from the untrained converter model,
• compare the output image fingerprint embedding with the second image fingerprint embedding associated with the respective image to calculate a loss, and
• backpropagate the loss through the untrained converter model to generate a trained converter model trained to convert a first image fingerprint embedding generated by the first image fingerprinting model to a second image fingerprint embedding compatible with other second image fingerprint embeddings generated by the second image fingerprinting model, and
• a conversion component comprising instructions that, upon execution by the one or more processors, cause the one or more processors to:
• access a plurality of stored first image fingerprint embeddings;
• provide each of the stored first image fingerprint embeddings as input to the trained converter model to generate a second plurality of second image fingerprint embeddings;
• and
• add the second plurality of second image fingerprint embeddings to an index comprising a third plurality of second image fingerprint embeddings generated by the second image fingerprinting model, wherein the index is used by an approximate nearest neighbors model.

A computer-readable storage device having stored thereon instructions that, upon execution by one or more processors, cause the one or more processors to:

• generate a plurality of first image fingerprint embeddings, wherein the instructions to generate comprise instructions that, upon execution by the one or more processors, cause the one or more processors to:
• provide each of a plurality of images as input to a first image fingerprinting model, wherein the first image fingerprinting model is trained to generate a first image fingerprint embedding comprising a multi-dimensional feature vector representing visual aspects of the input image, and
• receive each of the plurality of first image fingerprint embeddings as output from the first image fingerprinting model;
• generate a plurality of second image fingerprint embeddings, wherein the instructions to generate comprise instructions that, upon execution by the one or more processors, cause the one or more processors to:
• provide each of the plurality of images as input to a second image fingerprinting model, wherein the second image fingerprinting model is trained to generate a second image fingerprint embedding comprising a multi-dimensional feature vector representing visual aspects of the input image, and
• receive each of the plurality of second image fingerprint embeddings as output from the second image fingerprinting model, wherein the first image fingerprint embedding is different than the second image fingerprint embedding generated for each of the plurality of images;
• train a converter model to convert a first image fingerprint embedding to a second image fingerprint embedding compatible with other second image fingerprint embeddings, wherein the instructions to train comprise instructions that, upon execution by the one or more processors, cause the one or more processors to:
• for each image of the plurality of images:
• provide the first image fingerprint embedding associated with the respective image as input to the converter model,
• receive an output image fingerprint embedding from the converter model,
• compare the output image fingerprint embedding with the second image fingerprint embedding associated with the respective image to calculate a loss, and
• backpropagate the loss through the converter model;
• access a plurality of stored first image fingerprint embeddings;
• provide each of the stored first image fingerprint embeddings as input to the converter model to generate a second plurality of second image fingerprint embeddings;
• and
• add the second plurality of second image fingerprint embeddings to an index comprising a third plurality of second image fingerprint embeddings generated by the second image fingerprinting model, wherein the index is used by an approximate nearest neighbors model.