Product category optimization for image similarity searching of image-based listings in a network-based publication system

A computer-readable hardware storage device having embedded therein a set of instructions which, when executed by one or more processors of a computer, causes the computer to execute operations comprising:

• detecting a query image from a client machine, the query image generated by selection of a selectable icon at a user interface at the client machine, the selection causing configuration of the query image as a message in a transport layer transmitted by the client machine, the query image comprising an image of an item with predefined specified visual attributes, the item being of a given category;
• using a similarity search, search a database for a set of images of items of the given category, among a collection of images of items, that have visual attributes that are similar to the visual attributes of the query image;
• based on the searching, determining a set of images of items of the given category that have visual attributes that are similar to the visual attributes of the query image;
• ranking a plurality of images of the set of images of items of the given category that have visual attributes that are similar to the specified visual attributes by weighting the respective ones of visual attributes of the plurality of images with weights, the weights being calculated by at least one hardware processor correlating judgment scores comprising feedback signals received from human judges who, responsive to being asked to rank the similarity of various candidate images to a target image, provide the ranking as the feedback signals that indicate relative importance of individual ones of the predefined specified visual attributes to, the relative importance of individual ones of the predefined specified visual attributes differing from category to category;
• and
• presenting the ranked set of images to the client machine.

A system comprising:

• at least one computer processor configured to include:
• a detection module to detect a query image from a client machine, the query image generated by selection of a selectable icon at a user interface at the client machine, the selection causing configuration of the query image as a message in a transport layer transmitted by the client machine, the query image comprising an image of an item with predefined specified visual attributes, the item being of a given category;
• a similarity search module to search a database for a set of images of items of the given category, among a collection of images of items, that have visual attributes that are similar to the visual attributes of the query image and, based on the search, determine a set of images of items of the given category that have visual attributes that are similar to the specified visual attributes;
• and
• rank a plurality of images of the set of images of items of the given category that have visual attributes that are similar to the specified visual attributes by weighting some of the respective visual attributes of the plurality of images with weights, the weights calculated by at least one hardware processor correlating judgment scores comprising feedback signals received from human judges who, responsive to being asked to rank the similarity of various candidate images to a target image, provide the ranking as the feedback signals that indicate relative importance of individual ones of the predefined specified visual attributes to the human judges, the relative importance of individual ones of the predefined specified visual attributes differing from category to category;
• and
• a presentation module to present the ranked set of images to the client machine.