Enhanced search with morphed images
A search system performs item searches using morphed images. Given two or more input images of objects, the search system uses a generative model to generate one or more morphed images. Each morphed image shows an object of the object type of the input images. Additionally, the object in each morphed image combines visual characteristics of the objects in the input images. The morphed images may be presented to a user, who may select a particular morphed image for searching. A search is performed using a morphed image to identify items that are visually similar to the object in the morphed image. Search results for the items identified from the search are returned for presentation.
1. A computer-implemented method comprising:
receiving two or more images of objects of a type of object;
generating, using a generative model trained on images of objects of the type of object, one or more morphed images, each morphed image showing an object of the type of object combining visual characteristics of the objects from the received two or more images;
providing the one or more morphed images for presentation;
receiving input indicative of a selection of a first morphed image from the one or more morphed images;
performing a search on an item database using the selected first morphed image as search input to identify one or more search results;
providing the one or more search results for presentation;
receiving a selection of a first search result from the one or more search results;
generating, using the generative model, a further morphed image using the selected first morphed image and an image corresponding to the first search result;
performing a search on the item database using the further morphed image as search input to identify one or more further search results; and
providing the one or more further search results for presentation.
2. The computer-implemented method of claim 1 , wherein the received two or more images are received from a user device.
3. The computer-implemented method of claim 1 , wherein each object from the received two or more images comprises an item from the item database.
4. The computer-implemented method of claim 1 , wherein the generative model comprises a generative adversarial network trained on the images of objects of the type of object.
5. The computer-implemented method of claim 1 , wherein a plurality of morphed images are generated by the generative model, wherein a first morphed image from the plurality of morphed images shows a first object having a visual similarity closer to the object of a first received image, and wherein a second morphed image from the plurality of morphed images shows a second object having a visual similarity closer to the object of a second received image.
6. The computer-implemented method of claim 1 , wherein the search is performed on the item database using one or more additional search parameters selected from the following: a text input, a filter selection, and a user preference.
7. One or more computer storage media storing computer-useable instructions that, when used by a computing device, cause the computing device to perform operations, the operations comprising:
receiving two or more images of objects of a type of object;
generating, using a generative model trained on images of objects of the type of object, one or more morphed images, each morphed image showing an object of the type of object combining visual characteristics of the objects from the received two or more images;
providing the one or more morphed images for presentation;
receiving input indicative of a selection of a first morphed image from the one or more morphed images;
performing a search on an item database using the selected first morphed image as search input to identify one or more search results;
providing the one or more search results for presentation;
receiving a selection of a first search result from the one or more search results;
generating, using the generative model, a further morphed image using the selected first morphed image and an image corresponding to the first search result;
performing a search on the item database using the further morphed image as search input to identify one or more further search results; and
providing the one or more further search results for presentation.
8. The computer storage media of claim 7 , wherein the received two or more images are received from a user device.
9. The computer storage media of claim 7 , wherein each object from the received two or more images comprises an item from the item database.
10. The computer storage media of claim 7 , wherein the generative model comprises a generative adversarial network trained on the images of objects of the type of object.
11. The computer storage media of claim 7 , wherein a plurality of morphed images are generated by the generative model, wherein a first morphed image from the plurality of morphed images shows a first object having a visual similarity closer to the object of a first received image, and wherein a second morphed image from the plurality of morphed images shows a second object having a visual similarity closer to the object of a second received image.
12. The computer storage media of claim 7 , wherein the search is performed on the item database using one or more additional search parameters selected from the following: a text input, a filter selection, and a user preference.
13. A computer system comprising:
a processor; and
a computer storage medium storing computer-useable instructions that, when used by the processor, causes the computer system to perform operations comprising:
receiving two or more images of objects of a type of object;
generating, using a generative model trained on images of objects of the type of object, one or more morphed images, each morphed image showing an object of the type of object combining visual characteristics of the objects from the received two or more images;
providing the one or more morphed images for presentation;
receiving input indicative of a selection of a first morphed image from the one or more morphed images;
performing a search on an item database using the selected first morphed image as search input to identify one or more search results;
providing the one or more search results for presentation;
receiving a selection of a first search result from the one or more search results.
generating, using the generative model, a further morphed image using the selected first morphed image and an image corresponding to the first search result;
performing a search on the item database using the further morphed image as search input to identify one or more further search results; and
providing the one or more further search results for presentation.
14. The computer system of claim 13 , wherein the received two or more images are received from a user device.
15. The computer system of claim 13 , wherein each object from the received two or more images comprises an item from the item database.
16. The computer system of claim 13 , wherein the generative model comprises a generative adversarial network trained on the images of objects of the type of object.
17. The computer system of claim 13 , wherein a plurality of morphed images are generated by the generative model, wherein a first morphed image from the plurality of morphed images shows a first object having a visual similarity closer to the object of a first received image, and wherein a second morphed image from the plurality of morphed images shows a second object having a visual similarity closer to the object of a second received image.