Abstract:
A generative network may be learned in an adversarial setting with a goal of modifying synthetic data such that a discriminative network may not be able to reliably tell the difference between refined synthetic data and real data. The generative network and discriminative network may work together to learn how to produce more realistic synthetic data with reduced computational cost. The generative network may iteratively learn a function that synthetic data with a goal of generating refined synthetic data that is more difficult for the discriminative network to differentiate from real data, while the discriminative network may be configured to iteratively learn a function that classifies data as either synthetic or real. Over multiple iterations, the generative network may learn to refine the synthetic data to produce refined synthetic data on which other machine learning models may be trained.
Abstract:
Some embodiments provide a novel user interface (UI) tool that is a unified slider control, which includes multiple sliders that slide along a region. The region is a straight line in some embodiments, while it is an angular arc in other embodiments. In some embodiments, the unified slider control is used in a media editing application to allow a user to modify several different properties of the image by moving several different sliders along the region. Each slider is associated with a property of the image. A position of the slider in the region corresponds to a value of the property associated with the slider.
Abstract:
Some embodiments provide a warning prompt control module which adjustably controls the display of warning prompts for vehicle elements in a vehicle, based on a determined profile with which the occupant is associated. An occupant detected in the vehicle interior can be associated with a profile based on a sensor data representation of the occupant correlating with a sensor data representation included in the profile, accessing a profile from a device supporting the occupant, etc. A profile can include interaction history data which indicate historical interactions with warning prompts for a vehicle element. Displaying a warning prompt for a vehicle element can be adjustably controlled based on the interaction history data, included in a profile, which is associated with the vehicle element.
Abstract:
A system and method that receives and edits image data of an underwater scene in a digital image in order to remove undesirable tints from objects in the scene. In some embodiments, colors near the color of the water itself are protected to leave the water looking blue. Removing undesirable tints without removing the tint of the water itself results in images with more realistic coloring of people and objects in the scene, without eliminating the color cues (e.g., blue water) that indicate that the image is a photograph of an underwater scene.
Abstract:
A method and system for controlling multiple image editing controls using one master control. The system identifies various characteristics of an image being edited. The system generates, for each of multiple image editing controls, a relationship between the master control and the image editing control. The relationship is based on at least one of the identified characteristics of the image being edited. The relationship is different for different images when the different images have different characteristics, such as different average color component values at a particular percentile of pixels in the images.
Abstract:
A system and method that receives and edits image data of an underwater scene in a digital image in order to remove undesirable tints from objects in the scene. In some embodiments, colors near the color of the water itself are protected to leave the water looking blue. Removing undesirable tints without removing the tint of the water itself results in images with more realistic coloring of people and objects in the scene, without eliminating the color cues (e.g., blue water) that indicate that the image is a photograph of an underwater scene.
Abstract:
Some embodiments of the image editing and organizing application described herein provide a multi-stage automatic enhancement process. The process takes an input image and feeds it through multiple different enhancement operations. The multiple enhancement operations of some embodiments are carried out in a particular order. In some embodiments, the particular order starts with exposure adjustment, then a white balance adjustment, then a vibrancy adjustment, then a tonal response curve adjustment, then a shadow lift adjustment.
Abstract:
Some embodiments provide an image editing application with a novel color modification slider tool. In some embodiments, this tool adjusts one or more color attributes (e.g., one or more color channels) of different pixels differently when moved in one direction, and adjusts one or more color attributes of the same pixel differently when moved in another direction. The movement of this tool is in a straight line in some embodiments, while it is along an angular arc (e.g., along the circumference of a full or partial circle or elliptical shape) in other embodiments. This tool in some embodiments is a novel saturation tool that produces non-photorealistic effects by treating differently the highly saturated and lowly saturated pixels of an image. Specifically, in some embodiments, the saturation tool performs either (1) a positive de-saturation effect that de-saturates the low saturated pixels of an image while enhancing the saturation of the high saturated pixels, or (2) a negative de-saturation effect that de-saturates the high saturated pixels of the image while leaving intact or slightly modifying the saturation level of the lowly saturated pixels. For an original image that has some saturated pixels, both these effects produce grey looking images but the first effect has more saturated pixels and hence looks a bit more non-photorealistic because of it contains grey and colorful pixels.