Nvidia Corporation

Baidu USA LLC

Bryan Catanzaro

Andrew J Tao

Shubhabrata Sengupta

Kevin Shih

Robert Pottorff

Awni Hannun

Ryan Prenger

Sanjeev Satheesh

Erich Elsen

Gregory Diamos

Jesse Engel

Christopher Fougner

Karan Sapra

Rohan Badlani

Dario Amodei

Xu Han

Ting-Chun Wang

Wei Ping

Jun Zhan

Adam Coates

David Tarjan

Zhenyao Zhu

Chong Wang

Guilin Liu

Mike Chrzanowski

Andrew Ng

Adrian Lancucki

Jose Rafael Valle Gomes Da Costa

Carl Case

Jared Casper

Jingdong Chen

Dani Yogatama

Fitsum A Reda

Stuart F Oberman

Ming Yiu Siu

Ilyas Elkin

Shiqiu Liu

Zhiding Yu

Animesh Garg

Jarmo Rafael Lundén

Aysegul Dundar

Jarmo Lunden

Sharanyan Chetlur

Erich Eisen

Rajarshi Roy

Jonathan Raiman

Neel Kant

Saad Godil

José Rafael Valle Gomes Da Costa

João Felipe Santos

Zhekun Luo

Brandon Rowlett

Robert Kirby

Sengupta Shubhabrata

Apparatuses, systems, and techniques to process image frames. In at least one embodiment, one or more edges of one or more objects within one or more images are digitally enhanced before downsampling the one or more objects.

Edge-enhanced video frame blending

Disclosed are apparatuses, systems, and techniques that may use machine learning for implementing generative text-to-speech models. The techniques include identifying a mapping of speech characteristics (SC) on a target distribution of a latent variable using a non-linear transformation for at least a subset of the SC. Parameters of the non-linear transformation are determined using a neural network that approximates a statistics of the SC with a statistics predicted for the SC based on the identified mapping and the target distribution of the latent variable.

Normalizing flows with neural splines for high-quality speech synthesis

A neural network architecture is disclosed for performing image in-painting using partial convolution operations. The neural network processes an image and a corresponding mask that identifies holes in the image utilizing partial convolution operations, where the mask is used by the partial convolution operation to zero out coefficients of the convolution kernel corresponding to invalid pixel data for the holes. The mask is updated after each partial convolution operation is performed in an encoder section of the neural network. In one embodiment, the neural network is implemented using an encoder-decoder framework with skip links to forward representations of the features at different sections of the encoder to corresponding sections of the decoder.

Image in-painting for irregular holes using partial convolutions

Apparatuses, systems, and techniques are presented to generate images with one or more visual effects applied. In at least one embodiment, one or more visual effects are applied to one or more images having a resolution that is less than a first resolution and those visual effects approximated for one or more images having a resolution that is greater than or equal to the first resolution.

Image enhancement using one or more neural networks

Apparatuses, systems, and techniques are presented to generate one or more images. In at least one embodiment, two or more pixels from two or more images are blended based, at least in part, on a distance of the two or more pixels from a region of the two or more images, in which pixel colors are substantially similar.

Pixel blending for neural network-based image generation

Apparatuses, systems, and techniques for designing a data path circuit such as a parallel prefix circuit with reinforcement learning are described. A method can include receiving a first design state of a data path circuit, inputting the first design state of the data path circuit into a machine learning model, and performing reinforcement learning using the machine learning model to output a final design state of the data path circuit, wherein the final design state of the data path circuit has decreased area, power consumption and/or delay as compared to conventionally designed data path circuits.

Data path circuit design using reinforcement learning

Apparatuses, systems, and techniques for texture synthesis from small input textures in images using convolutional neural networks. In at least one embodiment, one or more convolutional layers are used in conjunction with one or more transposed convolution operations to generate a large textured output image from a small input textured image while preserving global features and texture, according to various novel techniques described herein.

Techniques to use a neural network to expand an image

Apparatuses, systems, and techniques to enhance video are disclosed. In at least one embodiment, one or more neural networks are used to create, from a first video, a second video having one or more additional video frames.

Video prediction using one or more neural networks

Generation of synthetic speech from an input text sequence may be difficult when durations of individual phonemes forming the input text sequence are unknown. A predominantly parallel process may model speech rhythm as a separate generative distribution such that phoneme duration may be sampled at inference. Additional information such as pitch or energy may also be sampled to provide improved diversity for synthetic speech generation.

Unsupervised alignment for text to speech synthesis using neural networks

Presented herein are embodiments of state-of-the-art speech recognition systems developed using end-to-end deep learning. In embodiments, the model architecture is significantly simpler than traditional speech systems, which rely on laboriously engineered processing pipelines; these traditional systems also tend to perform poorly when used in noisy environments. In contrast, embodiments of the system do not need hand-designed components to model background noise, reverberation, or speaker variation, but instead directly learn a function that is robust to such effects. Neither a phoneme dictionary, nor even the concept of a 'phoneme,' is needed. Embodiments include a well-optimized recurrent neural network (RNN) training system that can use multiple GPUs, as well as a set of novel data synthesis techniques that allows for a large amount of varied data for training to be efficiently obtained. Embodiments of the system can also handle challenging noisy environments better than widely used, state-of-the-art commercial speech systems.

Deep learning models for speech recognition

Systems and methods for a multi-core optimized Recurrent Neural Network (RNN) architecture are disclosed. The various architectures affect communication and synchronization operations according to the Multi-Bulk-Synchronous-Parallel (MBSP) model for a given processor. The resulting family of network architectures, referred to as MBSP-RNNs, perform similarly to a conventional RNNs having the same number of parameters, but are substantially more efficient when mapped onto a modern general purpose processor. Due to the large gain in computational efficiency, for a fixed computational budget, MBSP-RNNs outperform RNNs at applications such as end-to-end speech recognition.

Systems and methods for a multi-core optimized recurrent neural network

Disclosed are systems and methods that implement efficient engines for computation-intensive tasks such as neural network deployment. Various embodiments of the invention provide for high-throughput batching that increases throughput of streaming data in high-traffic applications, such as real-time speech transcription. In embodiments, throughput is increased by dynamically assembling into batches and processing together user requests that randomly arrive at unknown timing such that not all the data is present at once at the time of batching. Some embodiments allow for performing steaming classification using pre-processing. The gains in performance allow for more efficient use of a compute engine and drastically reduce the cost of deploying large neural networks at scale, while meeting strict application requirements and adding relatively little computational latency so as to maintain a satisfactory application experience.

Systems and methods for efficient neural network deployments

Presented herein are embodiments of state-of-the-art speech recognition systems developed using end-to-end deep learning. In embodiments, the model architecture is significantly simpler than traditional speech systems, which rely on laboriously engineered processing pipelines; these traditional systems also tend to perform poorly when used in noisy environments. In contrast, embodiments of the system do not need hand-designed components to model background noise, reverberation, or speaker variation, but instead directly learn a function that is robust to such effects. A phoneme dictionary, nor even the concept of a 'phoneme,' is needed. Embodiments include a well-optimized recurrent neural network (RNN) training system that can use multiple GPUs, as well as a set of novel data synthesis techniques that allows for a large amount of varied data for training to be efficiently obtained. Embodiments of the system can also handle challenging noisy environments better than widely used, state-of-the-art commercial speech systems.

Systems and methods for speech transcription

Growing community of inventors

Cupertino, CA, United States of America

Bryan Catanzaro