Nvidia提出了一种神经对话头视频合成AI模型，使视频会议的带宽效率提高了10倍
NVIDIA researchers introduce an AI system that generates a realistic talking-head video of a person using one source image 和 a driving video. The source image encodes an individual’s appearance, 和 the driving video directs motions in the resulting video.
研究人员提出了一种纯神经渲染方法，其中将谈话视频...
优步 Engineering Releases 霍罗沃德 v0.21: New Features Include Local Gradient Aggregation For TensorFlow v1 和 v2
Horovod announces its v0.21 和 brings many powerful new features to the 霍罗沃德 community, making training deep learning models faster 和 easier than ever before.
Horovod was open-sourced in 2017, 和 it has grown to become the standard solution for scaling deep learning training to hundreds of GPUs. 霍罗沃德 can reduce training times from days...
AlphaFold：DeepMind的AI系统具有重大突破，可预测蛋白质折叠
Google的DeepMind引入了AI系统AlphaFold， 可以准确预测蛋白质如何折叠成形状。
Proteins comprise strings of amino acids that are fundamental building parts of the human body. These are encoded into DNA 和 fold into multiple elaborate shapes that define their function. It is difficult to figure out the protein structure...
摩奈: An Open-Source Imaging Framework To Accelerate AI in Healthcare, Ready For Production Using 英伟达 Clara Application Framework
MONAI (the Medical Open Network for AI) is a PyTorch-based framework that contributes to AI development for healthcare issues. It is looking to address medical imaging issues with industry-specific data handling, high-performance training workflows, 和 reproducible reference implementations of state-of-the-art approaches. 摩奈 is ready for production as of April 2020, with the NVIDIA Clara 应用框架's upcoming release.
展示adversarial.js工具：一种在浏览器中的交互式，对神经网络的攻击性演示
东京大学研究生肯尼·宋（Kenny Song） adversarial.js，这是一个交互式工具，用于显示使用Tensorflow.js进行对抗攻击的功能。
There has been increasing attention to the threat that adversarial attacks cause to cybersecurity systems. Some adversarial attack examples include glasses that fool facial recognition systems 和 the stickers pasted on stop...
麦吉尔大学，Facebook&Mila推出MeDAL：用于医学缩写消除歧义的NLP预训练数据集，共计14M条
At EMNLP 2020 （自然语言处理中的经验方法）会议， 蒙特利尔的一个研究小组介绍了一个大型医学文本数据集，旨在改善医学缩写的歧义性。
Correct terminology 和 related deep learning models for various tasks have a significant role in medicine 和 healthcare. However, there has been a lack of publicly available pre-training data in this...
脸书 AI推出Linformer：一种新的变压器架构，可捕捉仇恨暴力的仇恨言论和内容
部署系统以检测充满语或故意拼写错误的仇恨言论一直是一个巨大的挑战。如今，功能最强大，最先进的语言理解系统使用具有数十亿或数万亿参数的大型Transformer模型。就像Facebook AI一样 RoBERTa and XLM-R, some new models make constant 和 repetitive advancements that are state-of-the-art, but with the help of even larger models...
脸书 AI探索Kotlin语言的自动区分系统
Differentiable programming enables programs to optimize themselves 和 is a part of 脸书 AI’s broader efforts to build additional advanced machine learning programming tools. 脸书 AI is developing an automatic differentiation system for the Kotlin programming language.
研究人员表示，这项工作将进一步探索软件2.0，在此软件可以有效地自我编写。那个...
艾伯塔大学的研究人员提出了U2- Net：一种新型的深层网络架构，该架构可通过显着物体检测生成逼真的铅笔肖像
In the computer vision field, the process of detection 和 segmentation of the most noticeable objects from natural scenes is known as salient object detection (SOD).
Most of the existing SOD networks have similar designs 和 leverage the depth features extracted by backbone networks like AlexNet, ResNet, ResNeXt, DenseNet, VGG, etc. These backbone networks...
加州大学伯克利分校的研究人员将AI用于静音语音的数字发声
Researchers at 加州大学伯克利分校 have developed an AI model that detects ‘silent speech.’ The model is based on digital voicing to predict words 和 generate synthetic speech. Electromyography (EMG), with electrodes located at the face 和 throat, is used to detect the silent speech.
研究人员断言，该模型可以为...提供许多应用程序。
