Microsoft Technology Licensing, LLC

Wolf Kienzle

Kenneth Paul Hinckley

Piotr Dollar

David Nister

Michael F Cohen

Neel Suresh Joshi

Michael A Toelle

Matthew Sammis Ashman

Ivan Stojiljkovic

Magdalena Vukosavljevic

Matthieu Tony Uyttendaele

Mladen Radojevic

Sing Bing Kang

Steven Mark Drucker

Kumar Hemachandra Chellapilla

Johannes P Kopf

Mudit Agrawal

Kenneth P Hinkley

Matthew S Ashman

A method of object detection includes receiving a first image taken from a first perspective by a first camera and receiving a second image taken from a second perspective, different from the first perspective, by a second camera. Each pixel in the first image is offset relative to a corresponding pixel in the second image by a predetermined offset distance resulting in offset first and second images. A particular pixel of the offset first image depicts a same object locus as a corresponding pixel in the offset second image only if the object locus is at an expected object-detection distance from the first and second cameras. The method includes recognizing that a target object is imaged by the particular pixel of the offset first image and the corresponding pixel of the offset second image.

Stereoscopic object detection leveraging expected object distance

Various technologies described herein pertain to creation of an output hyper-lapse video from an input video. Values indicative of overlaps between pairs of frames in the input video are computed. A value indicative of an overlap between a pair of frames can be computed based on a sparse set of points from each of the frames in the pair. Moreover, a subset of the frames from the input video are selected based on the values of the overlaps between the pairs of the frames in the input video and a target frame speed-up rate. Further, the output hyper-lapse video is generated based on the subset of the frames. The output hyper-lapse video can be generated without a remainder of the frames of the input video other than the subset of the frames.

Real-time hyper-lapse video creation via frame selection

A method of object detection includes receiving a first image taken by a first stereo camera, receiving a second image taken by a second stereo camera, and offsetting the first image relative to the second image by an offset distance selected such that each corresponding pixel of offset first and second images depict a same object locus if the object locus is at an assumed distance from the first and second stereo cameras. The method further includes locating a target object in the offset first and second images.

Stereoscopic object detection leveraging assumed distance

Described is a technology by which an image such as a stitched panorama is automatically cropped based upon predicted quality data with respect to filling missing pixels. The image may be completed, including by completing only those missing pixels that remain after cropping. Predicting quality data may be based on using restricted search spaces corresponding to the missing pixels. The crop is computed based upon the quality data, in which the crop is biased towards including original pixels and excluding predicted low quality pixels. Missing pixels are completed by using restricted search spaces to find replacement values for the missing pixels, and may use histogram matching for texture synthesis.

Image completion including automatic cropping

The description relates to a smart ring. In one example, the smart ring can be configured to be worn on a first segment of a finger of a user. The example smart ring can include at least one flexion sensor secured to the smart ring in a manner that can detect a distance between the at least one flexion sensor and a second segment of the finger. The example smart ring can also include an input component configured to analyze signals from the at least one flexion sensor to detect a pose of the finger.

Smart ring

A 'Stroke Untangler' composes handwritten messages from handwritten strokes representing overlapping letters or partial letter segments are drawn on a touchscreen device or touch-sensitive surface. These overlapping strokes are automatically untangled and then segmented and combined into one or more letters, words, or phrases. Advantageously, segmentation and composition is performed without requiring user gestures, timeouts, or other inputs to delimit characters within words, and without using handwriting recognition-based techniques to guide untangling and composing of the overlapping strokes to form characters. In other words, the user draws multiple overlapping strokes. Those strokes are then automatically segmented and combined into one or more corresponding characters. Text recognition of the resulting characters is then performed. Further, the segmentation and combination is performed in real-time, thereby enabling real-time rendering of the resulting characters in a user interface window. A related drawing mode enables entry of drawings in combination with the handwritten characters.

User interface for overlapping handwritten text input

Technologies pertaining to composing, displaying, and/or transmitting handwritten content through utilization of a touch-sensitive display screen of a mobile computing device are described herein. A user of the mobile computing device can set forth strokes on the touch-sensitive display screen, one on top of another, wherein such strokes correspond to different handwritten characters. Stroke segmentation can be undertaken to determine which strokes correspond to which characters in a handwritten sequence of characters.

Composition of handwritten messages on mobile computing devices

Technologies pertaining to object detection are described herein. A cascaded classifier executes over subwindows of an image in a plurality of stages. A crosstalk cascade is employed to reject subwindows as being candidates for including an object that is desirably detected, where the crosstalk cascade is a combination of multiple cascades.

Crosstalk cascades for use in object detection

The subject disclosure is directed towards a face detection technology in which image data is classified as being a non-face image or a face image. Image data is processed into an image pyramid. Features, comprising pixel pairs of the image pyramid, are provided to stages of a cascading classifier to remove sub-window candidates that are classified as non-face sub-windows within each stage. The face detection technology continues with one or more subsequent stages to output a result as to whether the image contains a face.

Real-time face detection using pixel pairs

The subject disclosure pertains to systems and methods for personalization of a recognizer. In general, recognizers can be used to classify input data. During personalization, a recognizer is provided with samples specific to a user, entity or format to improve performance for the specific user, entity or format. Biased regularization can be utilized during personalization to maintain recognizer performance for non-user specific input. In one aspect, regularization can be biased to the original parameters of the recognizer, such that the recognizer is not modified excessively during personalization.

Growing community of inventors

Bellevue, WA, United States of America

Wolf Kienzle