Abstract: A method to determine activity in a sequence of successively acquired images of a scene, comprises: acquiring the sequence of images; for each image in the sequence of images, forming a feature block of features extracted from the image and determining image specific information including a weighting for the image; normalizing the determined weightings to form a normalized weighting for each image in the sequence of images; for each image in the sequence of images, combining the associated normalized weighting and associated feature block to form a weighted feature block; passing a combination of the weighted feature blocks through a predictive module to determine an activity in the sequence of images; and outputting a result comprising the determined activity in the sequence of images.

Abstract: A method to determine activity in a sequence of successively acquired images of a scene, comprises: acquiring the sequence of images; for each image in the sequence of images, forming a feature block of features extracted from the image and determining image specific information including a weighting for the image; normalizing the determined weightings to form a normalized weighting for each image in the sequence of images; for each image in the sequence of images, combining the associated normalized weighting and associated feature block to form a weighted feature block; passing a combination of the weighted feature blocks through a predictive module to determine an activity in the sequence of images; and outputting a result comprising the determined activity in the sequence of images.

Abstract: The technology relates to tuning a data translation block (DTB) including a generator model and a discriminator model. One or more processors may be configured to receive training data including an image in a second domain. The image in the second domain may be transformed into a first domain with a generator model. The transformed image may be processed to determine one or more outputs with one or more deep neural networks (DNNs) trained to process data in the first domain. An original objective function for the DTB may be updated based on the one or more outputs. The generator and discriminator models may be trained to satisfy the updated objective function.

Abstract: Tracking units for facial features with advanced training for natural rendering of human faces in real-time are provided. An example device receives a video stream, and upon detecting a visual face, selects a 3D model from a comprehensive set of head orientation classes. The device determines modifications to the selected 3D model to describe the face, then projects a 2D model of tracking points of facial features based on the 3D model, and controls, actuates, or animates hardware based on the facial features tracking points. The device can switch among an example comprehensive set of 35 different head orientation classes for each video frame, based on suggestions computed from a previous video frame or from yaw and pitch angles of the visual head orientation. Each class of the comprehensive set is trained separately based on a respective collection of automatically marked images for that head orientation class.

Abstract: The technology relates to tuning a data translation block (DTB) including a generator model and a discriminator model. One or more processors may be configured to receive training data including an image in a second domain. The image in the second domain may be transformed into a first domain with a generator model. The transformed image may be processed to determine one or more outputs with one or more deep neural networks (DNNs) trained to process data in the first domain. An original objective function for the DTB may be updated based on the one or more outputs. The generator and discriminator models may be trained to satisfy the updated objective function.

Abstract: A method to determine activity in a sequence of successively acquired images of a scene, comprises: acquiring the sequence of images; for each image in the sequence of images, forming a feature block of features extracted from the image and determining image specific information including a weighting for the image; normalizing the determined weightings to form a normalized weighting for each image in the sequence of images; for each image in the sequence of images, combining the associated normalized weighting and associated feature block to form a weighted feature block; passing a combination of the weighted feature blocks through a predictive module to determine an activity in the sequence of images; and outputting a result comprising the determined activity in the sequence of images.

Abstract: The technology relates to tuning a data translation block (DTB) including a generator model and a discriminator model. One or more processors may be configured to receive training data including an image in a second domain. The image in the second domain may be transformed into a first domain with a generator model. The transformed image may be processed to determine one or more outputs with one or more deep neural networks (DNNs) trained to process data in the first domain. An original objective function for the DTB may be updated based on the one or more outputs. The generator and discriminator models may be trained to satisfy the updated objective function.

Abstract: The technology relates to tuning a data translation block (DTB) including a generator model and a discriminator model. One or more processors may be configured to receive training data including an image in a second domain. The image in the second domain may be transformed into a first domain with a generator model. The transformed image may be processed to determine one or more outputs with one or more deep neural networks (DNNs) trained to process data in the first domain. An original objective function for the DTB may be updated based on the one or more outputs. The generator and discriminator models may be trained to satisfy the updated objective function.

Abstract: Tracking units for facial features with advanced training for natural rendering of human faces in real-time are provided. An example device receives a video stream, and upon detecting a visual face, selects a 3D model from a comprehensive set of head orientation classes. The device determines modifications to the selected 3D model to describe the face, then projects a 2D model of tracking points of facial features based on the 3D model, and controls, actuates, or animates hardware based on the facial features tracking points. The device can switch among an example comprehensive set of 35 different head orientation classes for each video frame, based on suggestions computed from a previous video frame or from yaw and pitch angles of the visual head orientation. Each class of the comprehensive set is trained separately based on a respective collection of automatically marked images for that head orientation class.

Abstract: Tracking units for facial features with advanced training for natural rendering of human faces in real-time are provided. An example device receives a video stream, and upon detecting a visual face, selects a 3D model from a comprehensive set of head orientation classes. The device determines modifications to the selected 3D model to describe the face, then projects a 2D model of tracking points of facial features based on the 3D model, and controls, actuates, or animates hardware based on the facial features tracking points. The device can switch among an example comprehensive set of 35 different head orientation classes for each video frame, based on suggestions computed from a previous video frame or from yaw and pitch angles of the visual head orientation. Each class of the comprehensive set is trained separately based on a respective collection of automatically marked images for that head orientation class.

Abstract: Tracking units for facial features with advanced training for natural rendering of human faces in real-time are provided. An example device receives a video stream, and upon detecting a visual face, selects a 3D model from a comprehensive set of head orientation classes. The device determines modifications to the selected 3D model to describe the face, then projects a 2D model of tracking points of facial features based on the 3D model, and controls, actuates, or animates hardware based on the facial features tracking points. The device can switch among an example comprehensive set of 35 different head orientation classes for each video frame, based on suggestions computed from a previous video frame or from yaw and pitch angles of the visual head orientation. Each class of the comprehensive set is trained separately based on a respective collection of automatically marked images for that head orientation class.

Abstract: Techniques for identifying a person in a target image are described. According to one of the techniques, identifying a person in a target image involves displaying, within a graphical user interface, an image that depicts one or more faces. One or more faces are automatically detected within the image. A user provides input that selects a face of the one or more faces to be a currently-selected face. A set of images are selected from a collection of images, where the set of images includes images that closely match the currently-selected face. Concurrently with display of the currently-selected face, each image in the set of images is displayed. Within the graphical user interface, a control is provided. The control enables a user to select a target image from the set of images. In response to detecting that the user has selected a target image using the control, the currently-selected face is associated with a person to which the target image corresponds.

Abstract: Sub-regions within one or more face images are identified within a digital image, and enhanced by applying an artificial glint symmetrically and/or synchronously to image data corresponding to sub-regions of eyes within the face image. An enhanced face image is generated including an enhanced version of the face that includes certain original pixels in combination with pixels corresponding to the one or more eye regions of the face with the artificial glint.

Abstract: Techniques for identifying a person in a target image are described. According to one of the techniques, identifying a person in a target image involves displaying, within a graphical user interface, an image that depicts one or more faces. One or more faces are automatically detected within the image. A user provides input that selects a face of the one or more faces to be a currently-selected face. A set of images are selected from a collection of images, where the set of images includes images that closely match the currently-selected face. Concurrently with display of the currently-selected face, each image in the set of images is displayed. Within the graphical user interface, a control is provided. The control enables a user to select a target image from the set of images. In response to detecting that the user has selected a target image using the control, the currently-selected face is associated with a person to which the target image corresponds.

Abstract: Sub-regions within one or more face images are identified within a digital image, and enhanced by applying an artificial glint symmetrically and/or synchronously to image data corresponding to sub-regions of eyes within the face image. An enhanced face image is generated including an enhanced version of the face that includes certain original pixels in combination with pixels corresponding to the one or more eye regions of the face with the artificial glint.

Abstract: Techniques for identifying a person in a target image are described. According to one of the techniques, identifying a person in a target image involves displaying, within a graphical user interface, an image that depicts one or more faces. One or more faces are automatically detected within the image. A user provides input that selects a face of the one or more faces to be a currently-selected face. A set of images are selected from a collection of images, where the set of images includes images that closely match the currently-selected face. Concurrently with display of the currently-selected face, each image in the set of images is displayed. Within the graphical user interface, a control is provided. The control enables a user to select a target image from the set of images. In response to detecting that the user has selected a target image using the control, the currently-selected face is associated with a person to which the target image corresponds.

Abstract: Sub-regions within one or more face images are identified within a digital image, and enhanced by applying an artificial glint symmetrically and/or synchronously to image data corresponding to sub-regions of eyes within the face image. An enhanced face image is generated including an enhanced version of the face that includes certain original pixels in combination with pixels corresponding to the one or more eye regions of the face with the artificial glint.

Abstract: A method for providing improved foreground/background separation in a digital image of a scene is disclosed. The method comprises providing a first map comprising one or more regions provisionally defined as one of foreground or background within the digital image; and providing a subject profile corresponding to a region of interest of the digital image. The provisionally defined regions are compared with the subject profile to determine if any of the regions intersect with the profile region. The definition of one or more of the regions in the map is changed based on the comparison.

Abstract: A digital camera has an integral flash and stores and displays a digital image. Under certain conditions, a flash photograph taken with the camera may result in a red-eye phenomenon due to a reflection within an eye of a subject of the photograph. A digital apparatus has a red-eye filter which analyzes at least one partial face region identified within the digital image for the red-eye phenomenon and modifies the image to eliminate the red-eye phenomenon by changing the red area to black. The modification of the image is enabled when a photograph is taken under conditions indicative of the red-eye phenomenon. The modification is subject to anti-falsing analysis which further examines the area around the red-eye area for indicia of the eye of the subject. The detection and correction can be optimized for performance and quality by operating on subsample versions of the image when appropriate.

Abstract: A digital camera has an integral flash and stores and displays a digital image. Under certain conditions, a flash photograph taken with the camera may result in a red-eye phenomenon due to a reflection within an eye of a subject of the photograph. A digital apparatus has a red-eye filter which analyzes at least one partial face region identified within the digital image for the red-eye phenomenon and modifies the image to eliminate the red-eye phenomenon by changing the red area to black. The modification of the image is enabled when a photograph is taken under conditions indicative of the red-eye phenomenon. The modification is subject to anti-falsing analysis which further examines the area around the red-eye area for indicia of the eye of the subject. The detection and correction can be optimized for performance and quality by operating on subsample versions of the image when appropriate.