Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for scoring hash functions. In one aspect, a method includes computing one or more first performance indicators for an objective function computed on each of one or more sets of input data elements using a set of hash functions. A first overall performance indicator is computed using each of the computed performance indicators. The candidate hash function is added to the set of hash functions to generate a second set of hash functions. Second performance indicators are computed for the objective function computed on each of the sets of input data elements using the second set of hash functions. A second overall performance indicator is computed using each of the computed second performance indicators, and a score is computed for the candidate hash function using the first overall performance indicator and the second overall performance indicator.

Abstract: Examples of methods and systems for creating swivel views from handheld video are described. In some examples, a method may be performed by a handheld device to receive or capture a video of a target object and the video may include a plurality of frames and content of the target object from a plurality of viewpoints. The device may determine one or more approximately corresponding frames of the video including content of the target object from a substantially matching viewpoint and may align the approximately corresponding frames of the video based on one or more feature points of the target object to generate an aligned video. The device may provide sampled frames from multiple viewpoints from the aligned video, configured for viewing the target object in a rotatable manner, such as in a swivel view format.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an image ranking model to rank images based on hashes of their contents using a lookup table. An image training set is received. An image ranking model is trained with the training set by generating an image hash for each image of the ordered pair of images based on one or more features extracted from the image, computing a first score for a first image hash of a first image of the pair and a second score for a second image hash of a second image of the pair using the image ranking model, determining whether to update the image ranking model based on the first score and the second score, and updating the image ranking model using an update value based on the first score and the second score.

Abstract: Systems and methods facilitating random number generation of hashes for video and/or audio are provided. In one embodiment, a system can include: a memory, and a microprocessor that executes computer executable components. The components can include a weighted distribution generation component that can generate a sampling distribution of a weighted combination of uniform distributions, and obtain a sample value from the sampling distribution. In one embodiment, horizontal regions of substantially equal area can be identified. The sample value can be obtained by selecting one of the horizontal regions, and uniformly selecting a coordinate from the horizontal region. The coordinate can correspond to a value on a horizontal axis of the sampling distribution, and the value can be equal to a sample value. The sample value can be employed to compute a hash employed in video and/or audio fingerprinting and/or in computing image descriptors for video.

Abstract: Systems and methods for quantizing a local descriptor in video fingerprinting applications are provided. In one or more embodiments, local features of a video are extracted and characterized by a set of feature dimensions. The feature dimensions are then quantized to yield a quantized local descriptor for the video. To introduce a degree of pseudorandom variation in the quantization grids, a cascaded random quantization technique is employed to quantize the dimensions, wherein a quantized value for a given dimension is used to quantize a next dimension in sequence.

Abstract: A system and method is provided that determines whether objects in one image are visually similar to objects in another image by replacing the images' backgrounds with other images, such as a solid color or an image with texture, and comparing the resulting histograms.

Abstract: A computer-implemented method can include receiving training data that includes a set of non-matching pairs and a set of matching pairs. The method can further include calculating a non-matching collision probability for each non-matching pair of the set of non-matching pairs and a matching collision probability for each matching pair of the set of matching pairs. The method can also include generating a machine learning model that includes a first threshold and a second threshold. An unknown item and a particular known item are classified as not matching when their collision probability is less than the first threshold, and as matching when their collision probability is greater than the second threshold. The first threshold and the second threshold can be selected based on a minimization of errors in classification of matching and non-matching pairs in the training data, and a maximization of a retrieval efficiency metric.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining image search results. One of the methods includes generating a plurality of feature vectors for each image in a collection of images, wherein each feature vector is associated with an image tile of an image, wherein each feature vector corresponds to one of a plurality of predetermined visual words. All images in the collection of images that share at least a threshold number of matching visual words associated with matching image tiles are classified as near-duplicate images.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an image ranking model to rank images based on hashes of their contents using a lookup table. An image training set is received. An image ranking model is trained with the training set by generating an image hash for each image of the ordered pair of images based on one or more features extracted from the image, computing a first score for a first image hash of a first image of the pair and a second score for a second image hash of a second image of the pair using the image ranking model, determining whether to update the image ranking model based on the first score and the second score, and updating the image ranking model using an update value based on the first score and the second score.

Abstract: A system and method is provided that determines whether objects in one image are visually similar to objects in another image by replacing the images' backgrounds with other images, such as a solid color or an image with texture, and comparing the resulting histograms.

Abstract: A method and apparatus are provided for performing an image search based on a search query having a portion P1 and a portion P2. Based on the first search query, a second search query is generated that includes a portion P3 and the portion P2 such that the second search query is broader in scope than the first search query, while still retaining the portion P2 of the first query. A first image search is then performed for the first search query to obtain a first set of search results and a second image search is performed for the second search query to obtain a second set of search results. Consequently, an image from the first set of search results is selected for presentation to a user, wherein the selection is based on content of the second set of search results.

Abstract: A method, computer program product, and computer system for identifying a first portion of a facial image in a first image, wherein the first portion includes noise. A corresponding portion of the facial image is identified in a second image, wherein the corresponding portion includes less noise than the first portion. One or more filter parameters of the first portion are determined based upon, at least in part, the first portion and the corresponding portion. At least a portion of the noise from the first portion is smoothed based upon, at least in part, the one or more filter parameters. At least a portion of face specific details from the corresponding portion is added to the first portion.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating image search results. One of the methods includes receiving first image search results responsive to a text query, each first image search result associated with a respective first score indicating a relevance of an image represented by the first image search result to the text query. Second image search results responsive to a query image are received, each second image search result associated with a respective second score indicating a measure of similarity between an image represented by the second image search result and the query image. A set of final image search results is selected including combining first scores and second scores of the selected first image search results. The final image search results are ordered by similarity to the query image.

Abstract: Image similarity operations are performed in which a seed image is analyzed, and a set of semantic classifications are determined from analyzing the seed image. The set of semantic classifications can include multiple positive semantic classifications. A distance measure is determined that is specific to the set of semantic classifications. The seed image is compared to a collection of images using the distance measure. A set of similar images is determined from comparing the seed image to the collection of images.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating image search results. One of the methods includes receiving first image search results responsive to a text query, each first image search result associated with a respective first score indicating a relevance of an image represented by the first image search result to the text query. Second image search results responsive to a query image are received, each second image search result associated with a respective second score indicating a measure of similarity between an image represented by the second image search result and the query image. A set of final image search results is selected including combining first scores and second scores of the selected first image search results. The final image search results are ordered by similarity to the query image.

Abstract: A method, computer program product, and computer system for identifying a first portion of a facial image in a first image, wherein the first portion includes noise. A corresponding portion of the facial image is identified in a second image, wherein the corresponding portion includes less noise than the first portion. One or more filter parameters of the first portion are determined based upon, at least in part, the first portion and the corresponding portion. At least a portion of the noise from the first portion is smoothed based upon, at least in part, the one or more filter parameters. At least a portion of face specific details from the corresponding portion is added to the first portion.

Abstract: A robust hashing method is applied to media data (e.g., video, image, and/or audio data), producing a hash output that is robust with respect to at least one attribute of the media data. A histogram is generated for the media data and the histogram is hashed using a weighted hashing procedure. The histogram can be derived from a plurality of randomized versions of the media file, each randomized version of the media file altered to a random extent with respect to the attribute. The histogram can also be derived from a plurality of feature descriptors computed for the media data that are coarsely encoded with respect to the attribute. The weighted hashing procedure includes assigning a weight to components of the histogram and applying a plurality of hash functions to a number of versions of each component, the number of versions based on the assigned weight.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training scoring models. One method includes storing data identifying a plurality of positive and a plurality of negative training images for a query. The method further includes selecting a first image from either the positive group of images or the negative group of images, and applying a scoring model to the first image. The method further includes selecting a plurality of candidate images from the other group of images, applying the scoring model to each of the candidate images, and then selecting a second image from the candidate images according to scores for the images. The method further includes determining that the scores for the first image and the second image fail to satisfy a criterion, updating the scoring model, and storing the updated scoring model.

Abstract: Systems and methods are provided herein relating to audio matching. A compact digest can be generated based on sets of triples, where triples are groupings of three interest points that meet threshold criteria. The compact digest can be used in identifying a potential audio match. A full digest can then be used in verifying the potential match. By using a compact digest to perform audio matching, the audio matching system can be scaled to encompass millions or billions of reference audio samples while still using the full digest to maintain accuracy.

Abstract: A combined fingerprint is generated for a video that can match two near-identical videos that differ only in their aspect ratios or formats. A transformation strategy is selected by selecting a first and a second aspect correction method. A first transformed video is generated by applying the first aspect correction method to the video. A second transformed video is generated by applying the second aspect correction method to the video. A first fingerprint is generated using the first transformed video. A second fingerprint is generated using the second transformed video. The combined fingerprint is generated by combining the first half of the first fingerprint with the second half of the second fingerprint.