Abstract: A media edit point selection process can include a media editing software application programmatically converting speech to text and storing a timestamp-to-text map. The map correlates text corresponding to speech extracted from an audio track for the media clip to timestamps for the media clip. The timestamps correspond to words and some gaps in the speech from the audio track. The probability of identified gaps corresponding to a grammatical pause by the speaker is determined using the timestamp-to-text map and a semantic model. Potential edit points corresponding to grammatical pauses in the speech are stored for display or for additional use by the media editing software application. Text can optionally be displayed to a user during media editing.

Abstract: A system identifies a video comprising frames associated with content tags. The system detects features for each frame of the video. The system identifies, based on the detected features, scenes of the video. The system determines, for each frame for each scene, a frame score that indicates a number of content tags that match the other frames within the scene. The system selects, for each scene, a set of key frames that represent the scene based on the determined frame scores. The system receives a search query comprising a keyword. The system generates, for display, search results responsive to the search query including a dynamic preview of the video. The dynamic preview comprises an arrangement of frames of the video corresponding to each scene of the video. Each of the arrangement of frames is selected from the selected set of key frames representing the respective scene of the video.

Abstract: A media edit point selection process can include a media editing software application programmatically converting speech to text and storing a timestamp-to-text map. The map correlates text corresponding to speech extracted from an audio track for the media clip to timestamps for the media clip. The timestamps correspond to words and some gaps in the speech from the audio track. The probability of identified gaps corresponding to a grammatical pause by the speaker is determined using the timestamp-to-text map and a semantic model. Potential edit points corresponding to grammatical pauses in the speech are stored for display or for additional use by the media editing software application. Text can optionally be displayed to a user during media editing.

Abstract: Methods, systems, and non-transitory computer readable media are disclosed for generating a composite image comprising objects in positions from two or more different digital images. In one or more embodiments, the disclosed system receives a sequence of images and identifies objects within the sequence of images. In one example, the disclosed system determines a target position for a first object based on detecting user selection of the first object in the target position from a first image. The disclosed system can generate a fixed object image comprising the first object in the target position. The disclosed system can generate preview images comprising the fixed object image with the second object sequencing through a plurality of positions as seen in the sequence of images. Based on a second user selection of a desired preview image, the disclosed system can generate the composite image.

Abstract: Embodiments of the present invention are directed towards providing contextual tags for an image based on a contextual analysis of associated images captured in the same environment as the image. To determine contextual tags, content tags can be determined for images. The determined content tags can be associated with categories based on a contextual classification of the content tags. These associated content tags can then be designated as contextual tags for a respective category. To associate these contextual tags with the images, the images can be iterated through based on how the images relate to the contextual tags. For instance, when an image is associated with a category, the contextual tags classified into that category can be assigned to that image.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for detecting changes to a point of interest between a selected version and a previous version of a digital image and providing a summary of the changes to the point of interest. For example, the disclosed system provides for display a selected version of a digital image and detects a point of interest within the selected version of the digital image. The disclosed system determines image modifications to the point of interest (e.g., tracks changes to the point of interest) to generate a summary of the image modifications. Moreover, the summary can indicate further information concerning image modifications applied to the selected point of interest, such as timestamp, editor, or author information.

Abstract: Techniques are provided herein for more efficiently storing images that have a common subject, such as product images that share the same product in the image. Each image undergoes an adaptive tiling procedure to split the image into a plurality of tiles, with each tile identifying a region of the image having pixels with the same content. The tiles across multiple images can then be clustered together and those tiles having identical content are removed. Once all duplicate tiles have been removed from the set of all tiles across the images, the tiles are once again clustered based on their encoding scheme and certain encoding parameters. Tiles within each cluster are compressed using the best compression technique for the tiles in each corresponding cluster. By removing duplicative tile content between numerous images of the same subject, the total amount of data that needs to be stored is reduced.

Abstract: An image generation system generates images of objects under different lighting conditions. An image of an object and lighting conditions for an output image are received. The lighting conditions may specify, for instance, a location and/or color of one or more light sources. The image of the object is decomposed into a shading component and a reflectance component. A machine learning model takes the reflectance component and specified lighting conditions as input, and generates an output image of the object under the specified lighting conditions.

Abstract: In implementations of search input generation for an image search, a computing device can capture image data of an environment scene that includes multiple objects. The computing device implements a search input module that can detect the multiple objects in the image data, and initiate a display of a selectable indication for each of the multiple objects. The search input module can then determine a subject object from the detected multiple objects, and generate the subject object as the search input for the image search.

Abstract: Techniques are provided for identifying objects (such as products within a physical store) within a captured video scene and indicating which of object in the captured scene matches a desired object requested by a user. The matching object is then displayed in an accentuated manner to the user in real-time (via augmented reality). Object identification is carried out via a multimodal methodology. Objects within the captured video scene are identified using a neural network trained to identify different types of objects. The identified objects can then be compared against a database of pre-stored images of the desired product to determine if a close match is found. Additionally, text on the identified objects is analyzed and compared to the text of the desired object. Based on either or both identification methods, the desired object is indicated to the user on their display, via an augmented reality graphic.

Abstract: Systems and techniques for a design-aware image search are described. The design-aware image search techniques described herein capture a design on an item to determine additional items with similar or matching designs. An image of the item is used to create an edge image of a design, and shape descriptors are generated describing features of the edge image. These shape descriptors are compared to shape descriptors associated with other images to locate images that have similar or matching designs as compared with the input image. The design-aware image search system may use these relationships to generate a search result with images or products having a design similar to the design on the input image.

Abstract: Described herein is a system and techniques for classification of subjects within image information. In some embodiments, a set of subjects may be identified within image data obtained at two different points in time. For each of the subjects in the set of subjects, facial landmark relationships may be assessed at the two different points in time to determine a difference in facial expression. That difference may be compared to a threshold value. Additionally, contours of each of the subjects in the set of subjects may be assessed at the two different points in time to determine a difference in body position. That difference may be compared to a different threshold value. Each of the subjects in the set of subjects may then be classified based on the comparison between the differences and the threshold values.

Abstract: Methods, systems, and non-transitory computer readable media are disclosed for accurately and efficiently identifying digital images that portray counterfeit products. The disclosed system can store key values and properties of authentic graphic elements (logos, trademarks, designs, etc.,). For example, the disclosed system can determine authentic graphic features comprising authentic shape features, authentic color features, and authentic text features corresponding with a graphic element. The disclosed system can further extract a graphic element from a digital image for comparison with the authentic graphic element. For instance, the disclosed system can extract graphic features of the graphic element, the graphic features comprising shape features, color features, and text features from the graphic element.

Abstract: Techniques are provided herein for more efficiently storing images that have a common subject, such as product images that share the same product in the image. Each image undergoes an adaptive tiling procedure to split the image into a plurality of tiles, with each tile identifying a region of the image having pixels with the same content. The tiles across multiple images can then be clustered together and those tiles having identical content are removed. Once all duplicate tiles have been removed from the set of all tiles across the images, the tiles are once again clustered based on their encoding scheme and certain encoding parameters. Tiles within each cluster are compressed using the best compression technique for the tiles in each corresponding cluster. By removing duplicative tile content between numerous images of the same subject, the total amount of data that needs to be stored is reduced.

Abstract: Some implementations include methods for communicating features of images to visually impaired users. An image to be displayed on a touch sensitive screen of a computing device may include one or more objects. Each of the one or more objects may be associated with a bounding box. A contact with the image may be detected via the touch sensitive screen. The contact may be determined to be within a bounding box associated with a first object of the one or more objects. Responsive to detecting the contact to be within the bounding box associated with the first object, a caption of the first object may be caused to become audible and the touch sensitive screen may be caused to vibrate based on a vibration pattern unique to the first object.

Abstract: The present disclosure describes systems, non-transitory computer-readable media, and methods that can generate contextual identifiers indicating context for frames of a video and utilize those contextual identifiers to generate translations of text corresponding to such video frames. By analyzing a digital video file, the disclosed systems can identify video frames corresponding to a scene and a term sequence corresponding to a subset of the video frames. Based on images features of the video frames corresponding to the scene, the disclosed systems can utilize a contextual neural network to generate a contextual identifier (e.g. a contextual tag) indicating context for the video frames. Based on the contextual identifier, the disclosed systems can subsequently apply a translation neural network to generate a translation of the term sequence from a source language to a target language. In some cases, the translation neural network also generates affinity scores for the translation.

Abstract: Systems and techniques for a design-aware image search are described. The design-aware image search techniques described herein capture a design on an item to determine additional items with similar or matching designs. An image of the item is used to create an edge image of a design, and shape descriptors are generated describing features of the edge image. These shape descriptors are compared to shape descriptors associated with other images to locate images that have similar or matching designs as compared with the input image. The design-aware image search system may uses these relationships to generate a search result with images or products having a design similar to the design on the input image.

Abstract: Digital image ordering based on object position and aesthetics is leveraged in a digital medium environment. According to various implementations, an image analysis system is implemented to identify visual objects in digital images and determine aesthetics attributes of the digital images. The digital images can then be arranged in way that prioritizes digital images that include relevant visual objects and that exhibit optimum visual aesthetics.

Abstract: Techniques are described for trace layer for replicating a source region of a digital image. In an implementation, a user leverages a content editing system to select a source region of a source image to be replicated and a target region of a target image to which portions of the source region are to be replicated. A trace layer is generated that is a visual representation of portions of the source region, and the trace layer is positioned on the target region of the target image. Further, the trace layer is generated based on a visibility factor such that the trace layer is at least partially transparent. The trace layer receives user interaction to select portions of the trace layer and visibility of the selected portions is modified to replicate corresponding portions of the source region to the target region.

Abstract: Techniques are provided herein for more efficiently storing images that have a common subject, such as product images that share the same product in the image. Each image undergoes an adaptive tiling procedure to split the image into a plurality of tiles, with each tile identifying a region of the image having pixels with the same content. The tiles across multiple images can then be clustered together and those tiles having identical content are removed. Once all duplicate tiles have been removed from the set of all tiles across the images, the tiles are once again clustered based on their encoding scheme and certain encoding parameters. Tiles within each cluster are compressed using the best compression technique for the tiles in each corresponding cluster. By removing duplicative tile content between numerous images of the same subject, the total amount of data that needs to be stored is reduced.