Shutterstock

Abstract Methods, systems, and storage media for refining search queries through interactive conversational exchange are disclosed. Exemplary implementations may receive a first search input from a user describing a desired asset such as an image. Aspects implementations may also interact with the user via a chat interface to solicit additional details about the desired asset in response to the first search input, construct a search query based on the first search input and the additional details solicited from the user, and submit the search query to a search service to retrieve relevant asset data for the user.

Abstract Methods and systems for predicting performance of creative content are disclosed. Exemplary implementations may: receive a collection of images; provide a context to a user; serially cause display of pairs of images on a computer interface; receive user responses indicating which image of each pair is preferred given the context; determine a resonance value for each image based on a number of times the user responses indicate each image is preferred when displayed in a pair of images; determine a confidence score for each image; generate one or more models for predicting image performance based on one or more of the resonance value and the confidence score for each image; receive a plurality of candidate images; determine, using at least one model, a first metric set for each candidate; and cause display of a listing of the candidate images, the listing including the first metric set for each candidate image.

Abstract As disclosed herein, a computer-implemented method for refining a description of a desired digital asset through interactive conversational exchange is provided. The computer-implemented method may include receiving, via a conversational user interface (UI), a first input from a user including a description of a desired digital asset. The computer-implemented method may include prompting the user to provide a second input including additional details about the desired digital asset. The computer-implemented method may include generating, based on the second input, a first refined description of the desired digital asset. The computer-implemented method may include providing the first refined description to a machine learning (ML) model to generate the desired digital asset. A system and a non-transitory computer-readable storage medium are also disclosed.

Abstract A method for recommending results to a user from a search query is provided. The method includes receiving, in a search engine, a search query for a media file from a user, identifying a style preference of the user associated with a one or more media file attributes, based on a user-related search history, selecting, from a database, a one or more media files based on the search query and the style preference of the user, determining a style preference score for the one or more media files based on the media file attributes, and recommending to the user a top ranked media file based on the style preference score. A system including a memory storing instructions and one or more processors to execute the instructions to cause the system to perform the above method is also provided.

Abstract A method for training a generative image model, including defining multiple sensitive categories and protected attributes associated with multiple training images, determining for a particular sensitive category a distribution of a protected attribute within the training images, and based on the distribution, calculating for each training image a corresponding image debiasing weight value associated with the protected attribute. The method further includes generating annotated training data including the training images, and for each training image, (1) the corresponding image debiasing weight value associated with the protected attribute and (2) a corresponding descriptive text caption. The method further includes performing a training process using the annotated training data to train a generative image model resulting in a trained model. A contribution of each image in the training images to an optimization loss of the training process is weighted during the training process using the corresponding image debiasing weight value.