Training image and text embedding models

A computer-implemented method, the method comprising:

• obtaining, by a computing system comprising one or more processors, a search query from a user computing system;
• processing, by the computing system, the search query with a text embedding model to generate a query embedding;
• obtaining, by the computing system, a plurality of image embeddings generated with an image embedding model, wherein the plurality of image embeddings were generated by processing a plurality of images with the image embedding model to generate the plurality of image embeddings, wherein the image embedding model was trained on a training dataset comprising a plurality of image-image training examples, wherein each image training example is descriptive of an image pair and selection data associated with the image pair;
• determining, by the computing system, one or more image search results based on the query embedding and the plurality of image embeddings, wherein the one or more image search results are based on a respective relevance score for each of the plurality of images, wherein the respective relevance score is determined based on a respective image embedding associated with a respective image of the plurality of images;
• and
• providing, by the computing system, the one or more image search results to the user computing system.

One or more non-transitory computer-readable media that collectively store instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations, the operations comprising:

• obtaining a search query from a user computing system, wherein the search query comprises a query image;
• processing the query image with an image embedding model to generate a query image embedding, wherein the image embedding model was trained on a training dataset comprising a plurality of image-image training examples, wherein each image training example is descriptive of an image pair and selection data associated with the image pair;
• obtaining a plurality of other embeddings generated with the image embedding model, wherein the plurality of other embeddings are associated with a plurality of images;
• determining one or more image search results based on the query image embedding and the plurality of other embeddings, wherein the one or more image search results are based on a respective relevance score for each of the plurality of images, wherein the respective relevance score is determined based on a respective embedding associated with a respective image of the plurality of images;
• and
• providing the one or more image search results to the user computing system.