Qualcomm Incorporated

Nvidia Corporation

Regan Blythe Towal

Somdeb Majumdar

Venkata Sreekanta Reddy Annapureddy

Aliakbar Aghamohammadi

Vikram Makam Gupta

David Jonathan Julian

David Nister

Raghuraman Krishnamoorthi

Jeffrey Alexander Levin

Ravindra Manohar Patwardhan

Victor Hokkiu Chan

Dexu Lin

Anthony Sarah

Daniel Hendricus Franciscus Dijkman

Jan Krzys Wegrzyn

Seyed Hesameddin Najafi Shoushtari

Ziad Asghar

Sachin Subhash Talathi

Vijay Chintalapudi

Carolina Parada

M Anthony Lewis

Shayegan Omidshafiei

Benjamin Samuel Schwartz

Maroof Mohammed Farooq

In various examples, a deep learning solution for path detection is implemented to generate a more abstract definition of a drivable path without reliance on explicit lane-markings&#x2014;by using a detection-based approach. Using approaches of the present disclosure, the identification of drivable paths may be possible in environments where conventional approaches are unreliable, or fail&#x2014;such as where lane markings do not exist or are occluded. The deep learning solution may generate outputs that represent geometries for one or more drivable paths in an environment and confidence values corresponding to path types or classes that the geometries correspond. These outputs may be directly useable by an autonomous vehicle&#x2014;such as an autonomous driving software stack&#x2014;with minimal post-processing.

Path detection for autonomous machines using deep neural networks

An apparatus for classifying an input includes a classifier and a feature extractor. The feature extractor is configured to generate a feature vector based on the input. The feature vector is also configured to set a number of elements of the feature vector to zero to produce a sparse feature vector. The sparse feature vector has the same dimensions as the feature vector generated by the feature extractor. However, the sparse feature vector includes fewer non-zero elements than the feature vector generated by the feature extractor. The feature vector is further configured to forward the sparse feature vector to the classifier to classify the input.

Generating a sparse feature vector for classification

A method of managing memory usage of a stored training set for classification includes calculating one or both of a first similarity metric and a second similarity metric. The first similarity metric is associated with a new training sample and existing training samples of a same class as the new training sample. The second similarity metric is associated with the new training sample and existing training samples of a different class than the new training sample. The method also includes selectively storing the new training sample in memory based on the first similarity metric, and/or the second similarity metric.

Feature selection for retraining classifiers

An apparatus for learning a rank of an artificial neural network is configured to decompose a weight tensor into a first weight tensor and a second weight tensor. A set of rank selection parameters are applied to the first weight tensor and the second weight tensor to truncate the rank of the first weight tensor and the second weight tensor. The set of rank selection parameters are updated simultaneously with the weight tensors by averaging updates calculated for each rank selection parameter of the set of rank selection parameters.

Learning a truncation rank of singular value decomposed matrices representing weight tensors in neural networks

A method of training a neural network model includes determining a specificity of multiple filters after a predetermined number of training iterations. The method also includes determining whether to continue training each filter of the multiple filters based at least in part on the specificity. The method further includes classifying an input based on features obtained by convolving the input with the trained filters.

Filter specificity as training criterion for neural networks

In general, techniques are described for autonomously performing default operations. A vehicle comprising a processor and a memory may be configured to perform the techniques. The processor may determine, while currently operating autonomously, a current context in which the vehicle is operating autonomously. The memory may store the current context. The processor may further determine, while currently operating autonomously and based on the current context, one of a plurality of operations to perform when the vehicle is unable to continue to operate autonomously, and unable to transfer control of the vehicle to an occupant, and perform the determined one of the plurality of operations.

Autonomously performing default operations based on current operating contexts

A method for classifying an object includes applying multiple confidence values to multiple objects. The method also includes determining a metric based on the multiple confidence values. The method further includes determining a classification of a first object from the multiple objects based on a knowledge-graph when the metric is above a threshold.

Knowledge-graph biased classification for data

Context-based priors are utilized in machine learning networks (e.g., neural networks) for detecting objects in images. The likely locations of objects are estimated based on context labels. A machine learning network identifies a context label of an entire image. Based on the context label, the network selects a set of likely regions for detecting objects of interest in the image.

Context-based priors for object detection in images

A method for guiding a robot equipped with a camera to facilitate three-dimensional (3D) reconstruction through sampling based planning includes recognizing and localizing an object in a two-dimensional (2D) image. The method also includes computing 3D depth maps for the localized object. A 3D object map is constructed from the depth maps. A sampling based structure is grown around the 3D object map and a cost is assigned to each edge of the sampling based structure. The sampling based structure may be searched to determine a lowest cost sequence of edges that may, in turn be used to guide the robot.

Object-focused active three-dimensional reconstruction

In general, techniques are described for ensuring occupant awareness in vehicles. A device comprising a processor may be configured to perform the techniques. The processor may be configured to determine an event which will change an operating characteristic of the vehicle, and determine an extent of time remaining until an estimated occurrence of the expected event. The processor may also be configured to determine a process by which to inform the occupant of an operating context of the vehicle based on the determined extent of time remaining until the estimated occurrence the expected event, and perform the process by which to inform the occupant of the operating context of the vehicle. The device may also include a memory configured to store the process.

Ensuring occupant awareness in vehicles

A method performed by an electronic device is described. The method includes obtaining an auxiliary product request or an auxiliary service request in conjunction with a vehicle request. The method also includes obtaining image data from an interior of at least one vehicle. The method further includes performing computer vision analysis on the image data, including processing the image data to classify one or more objects within a scene of the interior of the at least one vehicle. The method additionally includes determining, based on the classification, whether the one or more objects satisfy the auxiliary product request or the auxiliary service request within the interior of the at least one vehicle.

Systems and methods for customizing amenities in shared vehicles

A method of biasing a deep neural network includes determining whether an element has an increased probability of being present in an input to the network. The method also includes adjusting a bias of activation functions of neurons in the network to increase sensitivity to the element. In one configuration, the bias is adjusted without adjusting weights of the network. The method further includes adjusting an output of the network based on the biasing.

Incorporating top-down information in deep neural networks via the bias term

A method of motion planning includes observing an object from a first pose of an agent having a controllable camera. The method also includes determining one or more subsequent control inputs to move the agent and the camera to observe the object from at least one subsequent pose. The subsequent control input(s) are determined so as to minimize an expected enclosing measure of the object based on visual data collected from the camera. The method further includes controlling the agent and the camera based on the subsequent control input(s).

Active camera movement determination for object position and extent in three-dimensional space

An intelligent camera network cooperatively acquires images over a wireless network. The network automatically captures images based on a trigger. The trigger may include messages from other image capturing devices. A first image capture devices is triggered to acquire an image based on a message from at least one other image capture device.

Managing crowd sourced photography in a wireless network

A method of blink and averted gaze avoidance with a camera includes detecting an averted gaze of a subject and/or one or more closed eyes of the subject in response to receiving an input to actuate a camera shutter. The method also includes scheduling actuation of the camera shutter to a future estimated time period to capture an image of the subject when a gaze direction of the subject is centered on the camera and/or both eyes of the subject are open.

Growing community of inventors

San Diego, CA, United States of America

Regan Blythe Towal