Multi-image search

1 . A computer-implemented method for searching with multiple images, the method comprising:

obtaining, by a computing system comprising one or more processors, a first image and a second image;

processing, by the computing system, the first image with an embedding model to generate a first embedding, wherein the first embedding is associated with an embedding space;

processing, by the computing system, the second image with the embedding model to generate a second embedding, wherein the second embedding is associated with the embedding space;

determining, by the computing system, a third embedding based on the first embedding, the second embedding, and one or more learned distributions of the embedding space;

determining, by the computing system, one or more search results based on the third embedding, wherein the one or more search results are associated with one or more search result embeddings determined to be associated with the third embedding; and

providing, by the computing system, the one or more search results for display in a search results interface.

2 . The method of claim 1 , wherein the one or more learned distributions are determined based on determining a shared learned distribution between the first embedding and the second embedding, and wherein the embedding model is a machine-learned model trained to process images and generate embeddings descriptive of features in a respective image.

3 . The method of claim 1 , wherein the one or more learned distributions are associated with one or more learned properties.

4 . The method of claim 1 , wherein determining the third embedding comprises:

determining the first embedding and the second embedding are within a first learned distribution; and

determining other embeddings within the first learned distribution.

5 . The method of claim 1 , further comprising:

determining the first embedding of the first image is associated with features a user requests to be absent; and

determining the second embedding of the second image is associated with features a user is requesting to be identified during search.

6 . The method of claim 5 , further comprising:

determining the first embedding and the second embedding are within a first learned distribution;

determining the second embedding is within a second learned distribution; and

determining the first embedding is absent from the second learned distribution.

7 . The method of claim 6 , wherein determining the one or more search results comprises:

determining search results embeddings associated with a second learned distribution and absent from a first learned distribution.

8 . The method of claim 1 , wherein the third embedding is determined based on the first embedding, the second embedding, the one or more learned distributions, and one or more contexts.

9 . The method of claim 8 , wherein the one or more contexts comprises sources of the images.

10 . The method of claim 8 , wherein the one or more contexts comprises user text data.

11 . A computing system for searching with multiple images, the system comprising:

one or more processors; and

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

obtaining a search query, wherein the search query comprises a first image and a second image;

processing the first image and the second image to determine one or more shared attributes, wherein the first image is associated with the one or more shared attributes, and wherein the second image is associated with the one or more shared attributes;

processing the first image with an embedding model to generate a first embedding, wherein the first embedding is associated with an embedding space;

processing the second image with the embedding model to generate a second embedding, wherein the second embedding is associated with the embedding space;

determining the first embedding and the second embedding are associated with one or more learned distributions of the embedding space;

determining one or more search results based at least in part on the first embedding, the second embedding, and the one or more learned distributions, wherein the one or more search results are determined to be associated with the one or more learned distributions; and

providing the one or more search results in a search results interface.

12 . The system of claim 11 , wherein determining the one or more search results comprises:

determining one or more search result embeddings associated with the one or more learned distributions.

13 . The system of claim 11 , wherein the first embedding and the second embedding are within the one or more learned distributions.

14 . The system of claim 11 , wherein the one or more learned distributions are associated with a particular object type.

15 . The system of claim 11 , wherein the one or more learned distributions are associated with a particular image classification.

16 . The system of claim 11 , wherein determining the first embedding and the second embedding are associated with the one or more learned distributions of the embedding space comprise:

determining the first embedding and the second embedding are associated with a plurality of learned distributions of the embedding space; and

wherein the one or more search results are determined based on the plurality of learned distributions.

17 . 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 first image and a second image;

processing the first image with an embedding model to generate a first embedding, wherein the first embedding is associated with an embedding space;

processing the second image with the embedding model to generate a second embedding, wherein the second embedding is associated with the embedding space;

determining a third embedding based on the first embedding, the second embedding, and one or more learned distributions of the embedding space;

determining one or more search results based on the third embedding, wherein the one or more search results are associated with one or more search result embeddings determined to be associated with the third embedding; and

providing the one or more search results for display in a search results interface.

18 . The one or more non-transitory computer-readable media of claim 17 , wherein the operations further comprise:

processing the first image and the second image to determine one or more shared attributes.

19 . The one or more non-transitory computer-readable media of claim 18 , wherein the one or more search results are determined based on the third embedding and the one or more shared attributes.

20 . The one or more non-transitory computer-readable media of claim 17 , wherein the one or more search results are determined based on the third embedding and context data comprising user data and image metadata.