Qualcomm Incorporated

Upal Mahbub

Vasudev Bhaskaran

Wesley James Holland

Gokce Dane

Venkata Ravi Kiran Dayana

Rakesh Nattoji Rajaram

Micha Galor Gluskin

Ashar Ali

Samuel Sunarjo

Scott Barker

Ananthapadmanabhan Arasanipalai Kandhadai

Gerhard Reitmayr

Daniel James Guest

Rengaraj Thirupathi

Walker Curtis

Systems and techniques are provided for processing one or more frames. For example, a process can include obtaining a first plurality of frames associated with a first settings domain from an image capture system, wherein the first plurality of frames is captured prior to obtaining a capture input. The process can include obtaining a reference frame associated with a second settings domain from the image capture system, wherein the reference frame is captured proximate to obtaining the capture input. The process can include obtaining a second plurality of frames associated with the second settings domain from the image capture system, wherein the second plurality of frames is captured after the reference frame. The process can include, based on the reference frame, transforming at least a portion of the first plurality of frames to generate a transformed plurality of frames associated with the second settings domain.

Low-power fusion for negative shutter lag capture

Embodiments include systems and methods that may be performed by a processor of a computing device. Embodiments may be applied for keypoint detection in an image. In embodiments, the processor of the computing device may apply to an image a first-stage neural network to define and output a plurality of regions, apply to each of the plurality of regions a respective second-stage neural network to output a plurality of keypoints in each of the plurality of regions, and apply to the plurality of keypoints a third-stage neural network to determine a correction for each of the plurality of keypoints to provide corrected keypoints.

Multi-stage neural network process for keypoint detection in an image

Methods, systems, and apparatuses are provided to automatically detect objects within images. For example, an image capture device may capture an image, and may apply a trained neural network to the image to generate an object value and a class value for each of a plurality of portions of the image. Further, the image capture device may determine, for each of the plurality of image portions, a confidence value based on the object value and the class value corresponding to each image portion. The image capture device may also detect an object within at least one image portion based on the confidence values. Further, the image capture device may output a bounding box corresponding to the at least one image portion. The bounding box defines an area of the image that includes one or more objects.

Apparatus and methods for object detection using machine learning processes

Systems, methods, and non-transitory media are provided for predictive camera initialization. An example method can include obtain, from a first image capture device, image data depicting a scene; classify the scene based on the image data; based on the classification of the scene, predict a camera use event; and based on the predicted camera use event, adjust a power mode of at least one of the first image capture device and a second image capture device.

Camera initialization for reduced latency

Systems and techniques are provided for registering three-dimensional (3D) images to deformable models. An example method can include determining, based on an image of a target and associated depth information, a 3D mesh of the target; determining different sets of rotation and translation parameters based on modifications to rotation and translation parameters of the 3D mesh; generating, based on the different sets of rotation and translation parameters, different 3D meshes having different orientations, different poses, and/or different alignments relative to the target; determining different sets of model parameters associated with the different 3D meshes, based on modifications to the different sets of rotation and translation parameters; generating, based on the different sets of model parameters, different additional 3D meshes having different orientations, different poses, and/or different alignments relative to the target; and selecting a final 3D mesh of the target from the different additional 3D meshes.

Three-dimensional scan registration with deformable models

Systems and techniques are provided for modeling three-dimensional (3D) meshes using images. An example method can include receiving, via a neural network system, an image of a target and metadata associated with the image and/or a device that captured the image; determining, based on the image and metadata, first 3D mesh parameters of a first 3D mesh of the target, the first 3D mesh parameters and first 3D mesh corresponding to a first reference frame associated with the image and/or the device; and determining, based on the first 3D mesh parameters, second 3D mesh parameters for a second 3D mesh of the target, the second 3D mesh parameters and second 3D mesh corresponding to a second reference frame, the second reference frame including a 3D coordinate system of a real-world scene where the target is located.

Modeling objects from monocular camera outputs

A method is presented. The method includes determining a number of landmarks in an image comprising multiple pixels. The method also includes determining a number of channels for the image based on a function of the number of landmarks. The method further includes determining, for each one of the number of channels, a confidence of each pixel of the multiple pixels corresponding to a landmark. The method still further includes identifying the landmark in the image based on the confidence.

Compact encoded heat maps for keypoint detection networks

Systems, methods, and non-transitory media are provided for capturing a region of interest (ROI) with a multi-camera system. An example method can include initializing image sensors of an electronic device, each image sensor being initialized in a lower-power mode having a lower power consumption than a higher-power mode supported by one or more of the image sensors; obtaining images captured by the image sensors in the lower-power mode; determining, based on the images, that an ROI in a scene is within a field-of-view (FOV) of a first image sensor from the image sensors; based on determining that the ROI is within the FOV of the first image sensor, decreasing the lower-power mode of one or more second image sensors to a power-off mode or an additional lower-power mode having a lower power consumption than the lower-power mode; and capturing, using the first image sensor, an image of the ROI.

Region of interest capture for electronic devices

Embodiments include systems and methods for keypoint detection in an image. In embodiments, a processor of a computing device may apply to an image a first neural network that has been trained to define and output a plurality of regions. The processor may apply to each of the plurality of regions a respective second neural network to that has been trained to output a plurality of keypoints in each of the plurality of regions. The processor may apply to the plurality of keypoints a third neural network that has been trained to determine a correction for each of the plurality of keypoints to provide corrected keypoints suitable for the execution of an image processing function.

Growing community of inventors

San Diego, CA, United States of America

Upal Mahbub