Abstract: An example process includes: concurrently receiving an audio stream and a video stream; determining, based on a first portion of the audio stream received within a predetermined duration before a current time and a first portion of the video stream received within the predetermined duration before the current time, whether a visual attention of a user is directed to an electronic device while the user is speaking; and in accordance with a determination that the visual attention of the user is directed to the electronic device while the user is speaking: identifying a second portion of the audio stream to include user speech intended for the electronic device; initiating, by a digital assistant operating on the electronic device, a task based the second portion of the audio stream; and providing an output indicative of the initiated task.

Abstract: Systems and processes for training and compressing a convolutional neural network model include the use of quantization and layer fusion. Quantized training data is passed through a convolutional layer of a neural network model to generate convolutional results during a first iteration of training the neural network model. The convolutional results are passed through a batch normalization layer of the neural network model to update normalization parameters of the batch normalization layer. The convolutional layer is fused with the batch normalization layer to generate a first fused layer and the fused parameters of the fused layer are quantized. The quantized training data is passed through the fused layer using the quantized fused parameters to generate output data, which may be quantized for a subsequent layer in the training iteration.

Abstract: In one embodiment, a method includes providing, to a client system of a user, a user interface for display. The user interface may include a first set of options for selecting an artificial intelligence (AI) task for integrating into a user application, a second set of options for selecting one or more devices on which the user wants to deploy the selected AI task, and a third set of options for selecting one or more performance constraints specific to the selected devices. User specifications may be received based on user selections in the first, second, and third sets of options. A custom AI model may be generated based on the user specifications and sent to the client system of the user for integrating into the user application. The custom AI model once integrated may enable the user application to perform the selected AI task on the selected devices.

Abstract: In one embodiment, a method includes providing, to a client system of a user, a user interface for display. The user interface may include a first set of options for selecting an artificial intelligence (AI) task for integrating into a user application, a second set of options for selecting one or more devices on which the user wants to deploy the selected AI task, and a third set of options for selecting one or more performance constraints specific to the selected devices. User specifications may be received based on user selections in the first, second, and third sets of options. A custom AI model may be generated based on the user specifications and sent to the client system of the user for integrating into the user application. The custom AI model once integrated may enable the user application to perform the selected AI task on the selected devices.

Abstract: Systems and processes for training and compressing a convolutional neural network model include the use of quantization and layer fusion. Quantized training data is passed through a convolutional layer of a neural network model to generate convolutional results during a first iteration of training the neural network model. The convolutional results are passed through a batch normalization layer of the neural network model to update normalization parameters of the batch normalization layer. The convolutional layer is fused with the batch normalization layer to generate a first fused layer and the fused parameters of the fused layer are quantized. The quantized training data is passed through the fused layer using the quantized fused parameters to generate output data, which may be quantized for a subsequent layer in the training iteration.

Abstract: In one embodiment, a method includes providing, to a client system of a user, a user interface for display. The user interface may include a first set of options for selecting an artificial intelligence (AI) task for integrating into a user application, a second set of options for selecting one or more devices on which the user wants to deploy the selected AI task, and a third set of options for selecting one or more performance constraints specific to the selected devices. User specifications may be received based on user selections in the first, second, and third sets of options. A custom AI model may be generated based on the user specifications and sent to the client system of the user for integrating into the user application. The custom AI model once integrated may enable the user application to perform the selected AI task on the selected devices.

Abstract: In one embodiment, a plurality of images is received. The plurality of images are frames of a video file. A user requests for a thumbnail picture representative of the plurality of images. The plurality of images are filtered to obtain a set of images. The filtering can be based on a blurriness of the image, whether an image is near a scene transition, an amount of text depicted in the image, or a color level of the image. Valence scores may be determined for one or more of the images in the set of images. Valence scores are based on determining values of characteristics of an image that can predict user responses to the image. A first image from the set of images is selected based at least in part on the valence score of the first image. The first image is sent for display.

Abstract: Access is provided to optimal thumbnails that are extracted from a stream of video. Using a processing device configured with a model that incorporates preferences generated by the brain and behavior from the perception of visual images, the optimal thumbnail(s) for a given video is/are selected, stored and/or displayed.

Abstract: In one embodiment, a plurality of images is received. The plurality of images are frames of a video file. A user requests for a thumbnail picture representative of the plurality of images. The plurality of images are filtered to obtain a set of images. The filtering can be based on a blurriness of the image, whether an image is near a scene transition, an amount of text depicted in the image, or a color level of the image. Valence scores may be determined for one or more of the images in the set of images. Valence scores are based on determining values of characteristics of an image that can predict user responses to the image. A first image from the set of images is selected based at least in part on the valence score of the first image. The first image is sent for display.

Abstract: Access is provided to optimal thumbnails that are extracted from a stream of video. Using a processing device configured with a model that incorporates preferences generated by the brain and behavior from the perception of visual images, the optimal thumbnail(s) for a given video is/are selected, stored and/or displayed.

Abstract: A system and method provides techniques and analysis tools for measuring how individuals perceive and respond to valences and more subtle “micro-valences” present in stimuli. The invention includes a process that uses human neuroimaging and behavioral techniques to measure valence in order to predict how individuals will perceive and react to any stimulus designed to engage individuals, e.g., end users or consumers, including, but not limited to, products, brands, logos, packaging, banner ads, and advertisements and their subcomponents or features, such as shape, color, pattern, and material properties.