This week’s newsletter includes: a neural network for lip reading, a drone to defend your home, record breaking deep learning-powered JPEG compression, a style transfer implementation in TensorFlow, a self-driving car simulator in GTA V and more! 神经网络阅读唇语,保卫你住宅的无人机,使用深度学习压缩JPEG图片,一种基于TensorFlow的图片风格变换算法,在GTA V里的自动驾驶汽车。
LipNet is a ridiculously impressive LSTM recurrent network that attempts to read lips (imagine the possibilities!), achieving 93.4% accuracy on the GRID corpus, outperforming experienced human lipreaders and the previous 79.6% state-of-the-art accuracy. Read more… LipNet是一个不可思议的LSTM recurrent network可以读唇语。达到了93.4%的正确率,远远超过了之前的唇语阅读软件79.6%的正确率。
使用深度学习压缩JPEG
The gold standard of compression for JPEGs (JPEG 2000) has been beaten for the first time by an approach using deep learning, built by a team at Twitter. Prepare to see many more iterations on this idea! Read the paper… JPEG图片的黄金压缩标准 (JPEG 2000) 这么多年来第一次被打败。
Style transfer is a technique (popularized by Prisma) that utilizes deep learning to attempt to replicate an input photo using the style specified during training. Fast Style Transfer is an implementation using TensorFlow that can process an image in 100ms on a NVIDIA Titan X GPU. It even works on videos! Read more…
通过深度学习来实现特定风格的图片转换。
Deep learning-powered image upscaler Neural Enhance (built by must-follow Tweeter @alexjc) continues to improve! New models are demonstrating better performance when handling blur and noise. See examples here and here. Read more…
深度学习来提高图像的清晰度,可以处理模糊和图像噪音。
Why are just two car models used extensively in autonomous vehicle development? The Lexus RX450 and the Ford Fusion, with its sibling, the Lincoln MKZ (the car we work on at Udacity) appear to dominate the landscape. Read more about why…
My colleague David shared the news that Blackberry is partnering with Ford, and how transformational it could be for Blackberry. Could Blackberry turn their company around by building software for self-driving cars? Read more…
黑莓正在和福特合作研发自动驾驶软件。
Here’s how it works: The Sunflower Home Awareness System relies on the drone and a handful of in-ground smart lights to watch over your house. It detects motion, vibration and sound. Using machine learning, the system can distinguish between a human, a car and animals. Crazy! Read more…
通过一架无人机和一些智能照明灯来监测你的住宅。 基于机器学习,系统可以分辨出进入的是人,还是汽车或者是动物。
Having followed the work of @crizcraig on DeepDrive for quite a while, I was super excited to hear he has released instructions for anyone to get up and running with his self-driving car simulator in GTA V. It’s not a simple installation, and certainly not for the faint-hearted, but will be a ton of fun once running, promise! Read how to get up and running…
DeepDrive发布了一个任何人都能够在GTA V中模拟自动驾驶汽车的教程。
深度学习系列文章
1. 革命性的Deep Learning,深度学习十大应用!