公开(公告)号：US20210326583A1

公开(公告)日：2021-10-21

申请号：US17234787

申请日：2021-04-19

Applicant: Magic Leap, Inc.

Inventor： Daniel Jürg Donatsch ,  Srivignesh Rajendran

IPC: G06K9/00 ,  G06N3/08 ,  G06N3/04 ,  G06K9/62 ,  G06F3/01

Abstract: Techniques are disclosed for training a machine learning model to predict user expression. A plurality of images are received, each of the plurality of images containing at least a portion of a user's face. A plurality of values for a movement metric are calculated based on the plurality of images, each of the plurality of values for the movement metric being indicative of movement of the user's face. A plurality of values for an expression unit are calculated based on the plurality of values for the movement metric, each of the plurality of values for the expression unit corresponding to an extent to which the user's face is producing the expression unit. The machine learning model is trained using the plurality of images and the plurality of values for the expression unit.

公开(公告)号：US11048978B2

公开(公告)日：2021-06-29

申请号：US16185582

申请日：2018-11-09

Applicant: Magic Leap, Inc.

Inventor： Andrew Rabinovich ,  Vijay Badrinarayanan ,  Srivignesh Rajendran ,  Chen-Yu Lee

IPC: G06K9/62 ,  G06N3/04 ,  G06N3/08

Abstract: Methods and systems for meta-learning are described for automating learning of child tasks with a single neural network. The order in which tasks are learned by the neural network can affect performance of the network, and the meta-learning approach can use a task-level curriculum for multi-task training. The task-level curriculum can be learned by monitoring a trajectory of loss functions during training. The meta-learning approach can learn to adapt task loss balancing weights in the course of training to get improved performance on multiple tasks on real world datasets. Advantageously, learning to dynamically balance weights among different task losses can lead to superior performance over the use of static weights determined by expensive random searches or heuristics. Embodiments of the meta-learning approach can be used for computer vision tasks or natural language processing tasks, and the trained neural networks can be used by augmented or virtual reality devices.

公开(公告)号：US10963758B2

公开(公告)日：2021-03-30

申请号：US16366047

申请日：2019-03-27

Applicant: MAGIC LEAP, INC.

Inventor： Andrew Rabinovich ,  Vijay Badrinarayanan ,  Daniel Detone ,  Srivignesh Rajendran ,  Douglas Bertram Lee ,  Tomasz Malisiewicz

IPC: G06K9/00 ,  G06K9/66 ,  G06K9/62 ,  G06K9/46 ,  G06N3/04 ,  G06N3/08

Abstract: The present disclosure provides an improved approach to implement structure learning of neural networks by exploiting correlations in the data/problem the networks aim to solve. A greedy approach is described that finds bottlenecks of information gain from the bottom convolutional layers all the way to the fully connected layers. Rather than simply making the architecture deeper, additional computation and capacitance is only added where it is required.