Abstract: This disclosure describes one or more implementations of systems, non-transitory computer-readable media, and methods that train a named entity recognition (NER) model with noisy training data through a self-cleaning discriminator model. For example, the disclosed systems utilize a self-cleaning guided denoising framework to improve NER learning on noisy training data via a guidance training set. In one or more implementations, the disclosed systems utilize, within the denoising framework, an auxiliary discriminator model to correct noise in the noisy training data while training an NER model through the noisy training data. For example, while training the NER model to predict labels from the noisy training data, the disclosed systems utilize a discriminator model to detect noisy NER labels and reweight the noisy NER labels provided for training in the NER model.

Abstract: Techniques and systems are described for real time streamable page generation from a digital document. A page generator module generates individually streamable pages from a digital document in which metadata is written at a beginning of the streamable pages that is usable to control rendering of the page. Therefore, upon receipt of the streamable page by a rendering engine of a receiver device (e.g., a printer), the metadata is usable to render the objects included in the streamable page as received. The streamable pages are renderable by a receiver device as individual pages and/or portions of the pages are received.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for efficiently automating the preparation of accurate alpha matte animations and modified digital videos utilizing polarized light. For example, the disclosed systems obtain a plurality of polarized digital videos portraying an animation of a foreground subject backlit by a polarized light source. In some embodiments, the disclosed systems generate a plurality of corrected polarized digital videos by adjusting intensity values of the plurality of polarized digital videos based on intensity differences across the plurality of polarized digital videos. The disclosed systems generate an alpha matte animation comprising a plurality of alpha mattes from the plurality of corrected polarized digital videos or from the plurality of polarized digital videos.

Abstract: A method, apparatus, non-transitory computer readable medium, and system for document retrieval include obtaining a query and a document. A prompt generator generates a prompt for a reasoning model based on the query and the document. The reasoning model generates a reasoning result based on the prompt. In some cases, the reasoning result indicates that the document answers the query. A machine learning model provides the document in response to the query based on the reasoning result.

Abstract: In implementations of systems for visual reordering of partial vector objects, a computing device implements an order system to receive input data describing a region specified relative to a group of vector objects that includes a portion of a first vector object and a portion of second vector object. A visual order as between the portion of the first vector object and the portion of the second vector object within the region is determined. The order system computes a modified visual order as between the portion of the first vector object and the portion of the second vector object within the region based on the visual order. The order system generates the group of vector objects for display in a user interface using a render surface and a sentinel value to render pixels within the region in the modified visual order.

Abstract: Systems and methods for machine learning are described. The systems and methods include receiving target training data including a training image and ground truth label data for the training image, generating source network features for the training image using a source network trained on source training data, generating target network features for the training image using a target network, generating at least one attention map for training the target network based on the source network features and the target network features using a guided attention transfer network, and updating parameters of the target network based on the attention map and the ground truth label data.

Abstract: Multimodal digital audio generation techniques are described that leverage multimodal inputs such as a digital image and text to generate digital audio using machine learning. In one or more examples, a digital image and text are received. Image semantic information is extracted from the digital image using machine learning. Digital audio is generated using generative machine learning based on the text and the image semantic information. The digital audio is then rendered and output by a digital audio output device.

Abstract: In some embodiments, techniques for identifying email events generated by bot activity are provided. For example, a process may involve applying bot detection patterns to identify bot activity among email response events.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media for generating modified digital images by utilizing a patch match algorithm to generate nearest neighbor fields for a second digital image based on a nearest neighbor field associated with a first digital image. For example, the disclosed systems can identify a nearest neighbor field associated with a first digital image of a first resolution. Based on the nearest neighbor field of the first digital image, the disclosed systems can utilize a patch match algorithm to generate a nearest neighbor field for a second digital image of a second resolution larger than the first resolution. The disclosed systems can further generate a modified digital image by filling a target region of the second digital image utilizing the generated nearest neighbor field.

Abstract: Certain aspects and features of this disclosure relate to semantically-aware image extrapolation. In one example, an input image is segmented to produce an input segmentation map of object instances in the input image. An object generation network is used to generate an extrapolated semantic label map for an extrapolated image. The extrapolated semantic label map includes instances in the original image and instances that will appear in an outpainted region of the extrapolated image. A panoptic label map is derived from coordinates of output instances in the extrapolated image and used to identify partial instances and boundaries. Instance-aware context normalization is used to apply one or more characteristics from the input image to the outpainted region to maintain semantic continuity. The extrapolated image includes the original image and the outpainted region and can be rendered or stored for future use.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media that position objects across multiple perspectives within digital images to be equidistant. For instance, in some embodiments, the disclosed systems detect a user interaction for moving a first object within a first perspective of a digital image. Additionally, the disclosed systems extract a first distance between the first object within the first perspective and a joining edge between the first perspective and a second perspective of the digital image. The disclosed systems also extract a second distance between a second object within the second perspective of the digital image and the joining edge. Based on the first distance and the second distance, the disclosed systems modify the digital image by positioning the first object within the first perspective to be equidistant to the joining edge relative to the second object within the second perspective.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media that utilizes artificial intelligence to learn to recommend foreground object images for use in generating composite images based on geometry and/or lighting features. For instance, in one or more embodiments, the disclosed systems transform a foreground object image corresponding to a background image using at least one of a geometry transformation or a lighting transformation. The disclosed systems further generating predicted embeddings for the background image, the foreground object image, and the transformed foreground object image within a geometry-lighting-sensitive embedding space utilizing a geometry-lighting-aware neural network. Using a loss determined from the predicted embeddings, the disclosed systems update parameters of the geometry-lighting-aware neural network. The disclosed systems further provide a variety of efficient user interfaces for generating composite digital images.

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: The present disclosure is directed toward systems, methods, and non-transitory computer readable media that provide a graphical review interface for curating, reviewing, and approving digital design element variants utilizing multistate vector objects within a digital image. The disclosed systems generate, in response to a user interaction performed at a designer device, a digital image comprising a multistate vector object modifiable to depict variants of a graphical element within the digital image. Further, the disclosed systems provide the digital image for display within a variant review interface on a client device for reviewing the variants of the graphical element. Moreover, the disclosed systems receive, from the client device, an indication of a selected variant from among the variants indicated by the multistate vector object.

Abstract: Systems and methods for image segmentation are described. Embodiments of the present disclosure receive an image depicting an object; generate image features for the image by performing a convolutional self-attention operation that outputs a plurality of attention-weighted values for a convolutional kernel applied at a position of a sliding window on the image; and generate label data that identifies the object based on the image features.

Abstract: In one aspect, a query management module executing on a processor receives, from a large language model (LLM), a graph query generated by the LLM based on a natural language query (NLQ). A validation module identifies an error in the graph query. The query management module provides an indication of the error to the LLM. The query management module receives a modified graph query from the LLM. The validation module validates the modified graph query. Based on the validation of the modified graph query, the query management module executes the modified graph query against a knowledge graph to return a result as a response to the NLQ.

Abstract: Methods, systems, and non-transitory computer readable storage media are disclosed for automatically detecting and reconstructing patterns in digital images. The disclosed system determines structurally similar pixels of a digital image by comparing neighborhood descriptors that include the structural context for neighborhoods of the pixels. In response to identify structurally similar pixels of a digital image, the disclosed system utilizes non-maximum suppression to reduce the set of structurally similar pixels to collinear pixels within the digital image. Additionally, the disclosed system determines whether a group of structurally similar pixels define the boundaries of a pattern cell that forms a rectangular grid pattern within the digital image. The disclosed system also modifies a boundary of a detected pattern cell to include a human-perceived pattern object via a sliding window corresponding to the pattern cell.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media for panoptically guiding digital image inpainting utilizing a panoptic inpainting neural network. In some embodiments, the disclosed systems utilize a panoptic inpainting neural network to generate an inpainted digital image according to panoptic segmentation map that defines pixel regions corresponding to different panoptic labels. In some cases, the disclosed systems train a neural network utilizing a semantic discriminator that facilitates generation of digital images that are realistic while also conforming to a semantic segmentation. The disclosed systems generate and provide a panoptic inpainting interface to facilitate user interaction for inpainting digital images. In certain embodiments, the disclosed systems iteratively update an inpainted digital image based on changes to a panoptic segmentation map.

Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for utilizing machine learning models to generate refined depth maps of digital images utilizing digital segmentation masks. In particular, in one or more embodiments, the disclosed systems generate a depth map for a digital image utilizing a depth estimation machine learning model, determine a digital segmentation mask for the digital image, and generate a refined depth map from the depth map and the digital segmentation mask utilizing a depth refinement machine learning model. In some embodiments, the disclosed systems generate first and second intermediate depth maps using the digital segmentation mask and an inverse digital segmentation mask and merger the first and second intermediate depth maps to generate the refined depth map.

Abstract: Disclosed herein are various techniques for more precisely and reliably (a) positioning top and bottom border edges relative to textual content, (b) positioning left and right border edges relative to textual content, (c) positioning mixed edge borders relative to textual content, (d) positioning boundaries of a region of background shading that fall within borders of textual content, (e) positioning borders relative to textual content that spans columns, (f) positioning respective borders relative to discrete portions of textual content, (g) positioning collective borders relative to discrete, abutting portions of textual content, (h) applying stylized corner boundaries to a region of background shading, and (i) applying stylized corners to borders.