Abstract: Remote distribution of multiple neural network models to various client devices over a network can be implemented by identifying a native neural network and remotely converting the native neural network to a target neural network based on a given client device operating environment. The native neural network can be configured for execution using efficient parameters, and the target neural network can use less efficient but more precise parameters.

Abstract: Remote distribution of multiple neural network models to various client devices over a network can be implemented by identifying a native neural network and remotely converting the native neural network to a target neural network based on a given client device operating environment. The native neural network can be configured for execution using efficient parameters, and the target neural network can use less efficient but more precise parameters.

Abstract: Augmented reality (AR) and virtual reality (VR) environment enhancement using an eyewear device. The eyewear device includes an image capture system, a display system, and a position detection system. The image capture system and position detection system identify feature points within a point cloud that represents captured images of an environment. The display system presents image overlays to a user including enhancement graphics positioned at the feature points within the environment.

Abstract: An active depth detection system can generate a depth map from an image and user interaction data, such as a pair of clicks. The active depth detection system can be implemented as a recurrent neural network that can receive the user interaction data as runtime inputs after training. The active depth detection system can store the generated depth map for further processing, such as image manipulation or real-world object detection.

Abstract: Techniques for training a neural network having a plurality of computational layers with associated weights and activations for computational layers in fixed-point formats include determining an optimal fractional length for weights and activations for the computational layers; training a learned clipping-level with fixed-point quantization using a PACT process for the computational layers; and quantizing on effective weights that fuses a weight of a convolution layer with a weight and running variance from a batch normalization layer. A fractional length for weights of the computational layers is determined from current values of weights using the determined optimal fractional length for the weights of the computational layers. A fixed-point activation between adjacent computational layers is related using PACT quantization of the clipping-level and an activation fractional length from a node in a following computational layer.

Abstract: Among other things, embodiments of the present disclosure improve the functionality of computer software and systems by facilitating the automatic performance optimization of a software application based on the particular platform upon which the application runs. In some embodiments, the system can automatically choose a set of parameters or methods at run-time from a design space with pre-selected optimization methods and parameters (e.g., algorithms, software libraries, and/or hardware accelerators) for a specific task.

Abstract: Augmented reality experiences of a user wearing an electronic eyewear device are captured by at least one camera on a frame of the electronic eyewear device, the at least one camera having a field of view that is larger than a field of view of a display of the electronic eyewear device. An augmented reality feature or object is applied to the captured scene. A photo or video of the augmented reality scene is captured and a first portion of the captured photo or video is displayed in the display. The display is adjusted to display a second portion of the captured photo or video with the augmented reality features as the user moves the user's head to view the second portion of the captured photo or video. The captured photo or video may be transferred to another device for viewing the larger field of view augmented reality image.

Abstract: Remote distribution of multiple neural network models to various client devices over a network can be implemented by identifying a native neural network and remotely converting the native neural network to a target neural network based on a given client device operating environment. The native neural network can be configured for execution using efficient parameters, and the target neural network can use less efficient but more precise parameters.

Abstract: An active depth detection system can generate a depth map from an image and user interaction data, such as a pair of clicks. The active depth detection system can be implemented as a recurrent neural network that can receive the user interaction data as runtime inputs after training. The active depth detection system can store the generated depth map for further processing, such as image manipulation or real-world object detection.

Abstract: Remote distribution of multiple neural network models to various client devices over a network can be implemented by identifying a native neural network and remotely converting the native neural network to a target neural network based on a given client device operating environment. The native neural network can be configured for execution using efficient parameters, and the target neural network can use less efficient but more precise parameters.

Abstract: An active depth detection system can generate a depth map from an image and user interaction data, such as a pair of clicks. The active depth detection system can be implemented as a recurrent neural network that can receive the user interaction data as runtime inputs after training. The active depth detection system can store the generated depth map for further processing, such as image manipulation or real-world object detection.

Abstract: Remote distribution of multiple neural network models to various client devices over a network can be implemented by identifying a native neural network and remotely converting the native neural network to a target neural network based on a given client device operating environment. The native neural network can be configured for execution using efficient parameters, and the target neural network can use less efficient but more precise parameters.

Abstract: An active depth detection system can generate a depth map from an image and user interaction data, such as a pair of clicks. The active depth detection system can be implemented as a recurrent neural network that can receive the user interaction data as runtime inputs after training. The active depth detection system can store the generated depth map for further processing, such as image manipulation or real-world object detection.

Abstract: Remote distribution of multiple neural network models to various client devices over a network can be implemented by identifying a native neural network and remotely converting the native neural network to a target neural network based on a given client device operating environment. The native neural network can be configured for execution using efficient parameters, and the target neural network can use less efficient but more precise parameters.

Abstract: An active depth detection system can generate a depth map from an image and user interaction data, such as a pair of clicks. The active depth detection system can be implemented as a recurrent neural network that can receive the user interaction data as runtime inputs after training. The active depth detection system can store the generated depth map for further processing, such as image manipulation or real-world object detection.

Abstract: An active depth detection system can generate a depth map from an image and user interaction data, such as a pair of clicks. The active depth detection system can be implemented as a recurrent neural network that can receive the user interaction data as runtime inputs after training. The active depth detection system can store the generated depth map for further processing, such as image manipulation or real-world object detection.

Abstract: A venue system of a client device can submit a location request to a server, which returns multiple venues that are near the client device. The client device can use one or more machine learning schemes (e.g., convolutional neural networks) to determine that the client device is located in one of specific venues of the possible venues. The venue system can further select imagery for presentation based on the venue selection. The presentation may be published as ephemeral message on a network platform.

Abstract: A venue system of a client device can submit a location request to a server, which returns multiple venues that, are near the client device. The client device can use one or more machine learning schemes (e.g., convolutional neural networks) to determine that the client device is located in one of specific venues of the possible venues. The venue system can further select imagery for presentation based on the venue selection. The presentation may be published as ephemeral message on a network platform.