Abstract: Devices and techniques are generally described for virtual avatar generation using a user-submitted face image. In various examples, first image data including an image of a user face and hair may be received. A first machine learning model may predict a skin tone of skin of the user face in the first image data. Modified first image data may be generated by modifying pixels representing the user face using the predicted skin tone. First segmented image data representing the user face and at least a portion of the hair may be generated. Second image data representing a virtual avatar with a pre-defined head may be generated. First scaled image data may be generated by scaling the modified first segmented image data based at least in part on the virtual avatar. Third image data may be generate by rendering the first scaled image data on a body of the virtual avatar.

Abstract: A method of labeling training data includes inputting a plurality of unlabeled input data samples into each of a plurality of pre-trained neural networks and extracting a set of feature embeddings from multiple layer depths of each of the plurality of pre-trained neural networks. The method also includes generating a plurality of clusterings from the set of feature embeddings. The method also includes analyzing, by a processing device, the plurality of clusterings to identify a subset of the plurality of unlabeled input data samples that belong to a same unknown class. The method also includes assigning pseudo-labels to the subset of the plurality of unlabeled input data samples.

Abstract: A method of labeling data and training a model is provided. The method includes obtaining a set of images. The set of images includes a first subset and a second subset. The first subset is associated with a first set of labels. The method also includes generating a set of pseudo labels for the set of images and a second set of labels for the second subset based on the first subset, the second subset, a first machine learning model, and a domain adaption model. The method further includes generating second machine learning model. The second machine learning model is generated based on the set of images, the set of pseudo labels, the first set of labels, and the second set of labels. The second set of labels is updated based on one or more inferences generated by the second machine learning model.

Abstract: A system and method provide extractions of regions of interest from images hand annotated by reviewers by lifting annotations from images, filtering out text labels, reconstructing continuous closed boundaries, and marking the contained region.

Abstract: A multi-task text infilling system receives a digital image and identifies a region of interest of the image that contains original text. The system uses a machine learning model to determine, in parallel: a foreground image that includes the original text; a background image that omits the original text; and a binary mask that distinguishes foreground pixels from background pixels, The system receives a target mask that contains replacement text. The system then applies the target mask to blend the background image with the foreground layer image and yield a modified digital image that includes the replacement text and omits the original text.

Abstract: To replace text in a digital video image sequence, a system will process frames of the sequence to: define a region of interest (ROI) with original text in each of the frames; use the ROIs to select a reference frame from the sequence; select a target frame from the sequence; determine a transform function between the ROI of the reference frame and the ROI of the target frame; replace the original text in the ROI of the reference frame with replacement text to yield a modified reference frame ROI; and use the transform function to transform the modified reference frame ROI to a modified target frame ROI in which the original text is replaced with the replacement text. The system will then insert the modified target frame ROI into the target frame to produce a modified target frame. This process may repeat for other target frames of the sequence.

Abstract: Systems and methods are provided for generating custom garment patterns for producing clothing sized to fit any newly input human body representation. A set of reference sample bodies may be selected for which custom designed garment patterns are then obtained. Once reference patterns are obtained, a custom garment pattern for a particular new body may be automatically created by blending two or more of the reference patterns. For example, the neighboring reference bodies to the new body may be identified within a low-dimensional embedding space, and interpolation of garment parameters for the previously designed garment patterns for these reference bodies may be performed to produce a custom garment pattern.

Abstract: A method of labeling training data includes inputting a plurality of unlabeled input data samples into each of a plurality of pre-trained neural networks and extracting a set of feature embeddings from multiple layer depths of each of the plurality of pre-trained neural networks. The method also includes generating a plurality of clusterings from the set of feature embeddings. The method also includes analyzing, by a processing device, the plurality of clusterings to identify a subset of the plurality of unlabeled input data samples that belong to a same unknown class. The method also includes assigning pseudo-labels to the subset of the plurality of unlabeled input data samples.

Abstract: A method includes defining a model for a liquid while the liquid is positioned at least partially within a nozzle of a printer. The method also includes synthesizing video frames of the liquid using the model to produce synthetic video frames. The method also includes generating a labeled dataset that includes the synthetic video frames and corresponding model values. The method also includes receiving real video frames of the liquid while the liquid is positioned at least partially within the nozzle of the printer. The method also includes generating an inverse mapping from the real video frames to predicted model values using the labeled dataset. The method also includes reconstructing the liquid in the real video frames based at least partially upon the predicted model values.

Abstract: Embodiments described herein provide a system for generating semantically accurate synthetic images. During operation, the system generates a first synthetic image using a first artificial intelligence (AI) model and presents the first synthetic image in a user interface. The user interface allows a user to identify image units of the first synthetic image that are semantically irregular. The system then obtains semantic information for the semantically irregular image units from the user via the user interface and generates a second synthetic image using a second AI model based on the semantic information. The second synthetic image can be an improved image compared to the first synthetic image.

Abstract: A method of labeling a dataset includes inputting a testing set comprising a plurality of input data samples into a plurality of pre-trained machine learning models to generate a set of embeddings output by the plurality of pre-trained machine learning models. The method further includes performing an iterative cluster labeling algorithm that includes generating a plurality of clusterings from the set of embeddings, analyzing the plurality of clusterings to identify a target embedding with a highest duster quality, analyzing the target embedding to determine a compactness for each of the plurality of clusterings of the target embedding, and identifying a target cluster among the plurality of clusterings of the target embedding based on the compactness. The method further includes assigning pseudo-labels to the subset of the plurality of input data samples that are members of the target duster.

Abstract: A three-dimensional (3D) printer includes a nozzle and a camera configured to capture a real image or a real video of a liquid metal while the liquid metal is positioned at least partially within the nozzle. The 3D printer also includes a computing system configured to perform operations. The operations include generating a model of the liquid metal positioned at least partially within the nozzle. The operations also include generating a simulated image or a simulated video of the liquid metal positioned at least partially within the nozzle based at least partially upon the model. The operations also include generating a labeled dataset that comprises the simulated image or the simulated video and a first set of parameters. The operations also include reconstructing the liquid metal in the real image or the real video based at least partially upon the labeled dataset.

Abstract: A 3D printer includes a nozzle configured to jet a drop of liquid metal therethrough. The 3D printer also includes a light source configured to illuminate the drop with a pulse of light. A duration of the pulse of light is from about 0.0001 seconds to about 0.1 seconds. The 3D printer also includes a camera configured to capture an image, video, or both of the drop. The 3D printer also includes a computing system configured to detect the drop in the image, the video, or both. The computing system is also configured to characterize the drop after the drop is detected. Characterizing the drop includes determining a size of the drop, a location of the drop, or both in the image, the video, or both.

Abstract: At least one input image comprising curvilinear features is received. Latent representations of the input images are learned using a trained deep neural network. At least one boundary estimate is determined based on the latent representations. At least one segmentation estimate of the at least one input image is determined based on the latent representations. The at least one image is mapped to output segmentation maps based on the segmentation estimate and the at least one boundary estimate.

Abstract: A method of image annotation includes obtaining a candidate annotation map for an annotation task for an image from each of a set of annotation models wherein each of the candidate annotation maps includes suggested annotations for the image, receiving user selections or modifications of at least one of the suggested annotations from one or more of the candidate annotation maps, and generating a final annotation map based on the user selections or modifications from the one or more of the candidate annotation maps.

Abstract: A method operates a three-dimensional (3D) metal object manufacturing system to compensate for displacement errors that occur during object formation. In the method, image data of a metal object being formed by the 3D metal object manufacturing system is generated prior to completion of the metal object and compared to original 3D object design data of the object to identify one or more displacement errors. For the displacement errors outside a predetermined difference range, the method modifies machine-ready instructions for forming metal object layers not yet formed to compensate for the identified displacement errors and operates the 3D metal object manufacturing system using the modified machine-ready instructions.

Abstract: A system and method provide extractions of regions of interest from images hand annotated by reviewers by lifting annotations from images, filtering out text labels, reconstructing continuous closed boundaries, and marking the contained region.

Abstract: A method of labeling data and training a model is provided. The method includes obtaining a set of images. The set of images includes a first subset and a second subset. The first subset is associated with a first set of labels. The method also includes generating a set of pseudo labels for the set of images and a second set of labels for the second subset based on the first subset, the second subset, a first machine learning model, and a domain adaption model. The method further includes generating second machine learning model. The second machine learning model is generated based on the set of images, the set of pseudo labels, the first set of labels, and the second set of labels. The second set of labels is updated based on one or more inferences generated by the second machine learning model.

Abstract: A method includes illuminating a drop with a pulse of light from a light source. A duration of the pulse of light is from about 0.0001 seconds to about 0.1 seconds. The method also includes capturing an image, video, or both of the drop. The method also includes detecting the drop in the image, the video, or both. The method also includes characterizing the drop after the drop is detected. Characterizing the drop includes determining a size of the drop, a location of the drop, or both in the image, the video, or both.

Abstract: A method of image annotation includes selecting a plurality of annotation models related to an annotation task for an image, obtaining a candidate annotation map for the image from each of the plurality of annotation models, and selecting at least one of the candidate annotation maps to be displayed via a user interface, the candidate annotation maps comprising suggested annotations for the image. The method further includes receiving user selections or modifications of at least one of the suggested annotations from the candidate annotation map and generating a final annotation map based on the user selections or modifications.