We propose the first deep learning approach for exemplar-based local colorization. Given a reference color image, our convolutional neural network directly maps a grayscale image to an output colorized image. Rather than using hand-crafted rules as in traditional exemplar-based methods, our end-to-end colorization network learns how to select, propagate, and predict colors from the large-scale data. The approach performs robustly and generalizes well even when using reference images that are unrelated to the input grayscale image. More importantly, as opposed to other learning-based colorization methods, our network allows the user to achieve customizable results by simply feeding different references.

One issue might be that many people have moved to ALE & OpenAI's Gym interface for API/environment implementations, and Python for implementation language. Your C library makes Python sound like a very second-class citizen, which is discouraging, and C is increasingly disfavored for its complexity & low-level nature. Just to get started with this, one has to learn the'Cassandra POMDP format', whatever that is, and then deal with C rather than Python. Are there that many people who want to solve MDPs in a tabular form whose preferred language is C and love defining their models in Cassandra POMDP format? You also don't have any impressive use-cases or demos of things which one can do easily in AIToolbox which can't be done elsewhere as easily, or as fast, or at all - what gives me any confidence that this is really mature and I won't simply invest days into learning it only to discover some severe limitation which makes it useless for me?

Those who are working on data parallel distributed training for neural networks, you might have come across this paper in your life. This work solves the communication bottleneck in data parallel DNN training by reducing the data transmission size by sending only gradient values that crosses a threshold. I was fascinated by the ideas in the paper and wanted to quickly try it out. So I made a version of DGC on MXNet. I tried training ResNet-110 on CIFAR-10 with DGC.

Abstract: Automatic neural architecture design has shown its potential in discovering powerful neural network architectures. Existing methods, no matter based on reinforcement learning or evolutionary algorithms (EA), conduct architecture search in a discrete space, which is highly inefficient. In this paper, we propose a simple and efficient method to automatic neural architecture design based on continuous optimization. We call this new approach neural architecture optimization (NAO). There are three key components in our proposed approach: (1) An encoder embeds/maps neural network architectures into a continuous space.

Please send me a message if you've worked with Amazon and Google's Deep Learning object identification programs. I am having trouble in setting these up and running some tests for a paper that I need to complete and would love some help. Shoot me a message and I would be more than happy to explain further in detail.

The idea that artificial intelligence can compose music is scary for a lot of people, including me. But music-making AI software has advanced so far in the past few years that it's no longer a frightening novelty; it's a viable tool that can and is being used by producers to help in the creative process. This raises the question: could artificial intelligence one day replace musicians? For the second episode of The Future of Music, I went to LA to visit the offices of AI platform Amper Music and the home of Taryn Southern, a pop artist who is working with Amper and other AI platforms to co-produce her debut album I AM AI. Using AI as a tool to make music or aid musicians has been in practice for quite some time.

Mattel has succeeded in (almost) bringing dinosaurs back from extinction. The toy giant has revealed its $250 'Alpha Training Blue' robot velociraptor toy that users can play with, pet and train with a remote controller, just like Chris Pratt in the'Jurassic World' movies. It walks, squawks and snarls just like a real raptor and is packed with loads of tricks so that users never get bored. Mattel has revealed its $250 'Alpha Training Blue' robot velociraptor toy that users can play with, pet and train with a remote controller, just like Chris Pratt in the'Jurassic World' movies Mattel said Thursday its Alpha Training Blue robo-raptor is now available for pre-order on Amazon. The toy will ship in October.

But the applications of AI in media are not just limited to content personalization. Media teams have to deal with manual processes for everything - from tagging the media to creating multilingual subtitles. But recent advances in AI are automating many of these tasks. Developments in computer vision, speech to text and natural language processing algorithms are changing the face of media creation, distribution and most importantly, media consumption. Voice is the most natural way for people to communicate.

Think of it as style transfer for dancing, a deep learning based algorithm that can convincingly show a real person mirroring the moves of their favorite dancers. The work, developed by a team of researchers from the University of California Berkeley, allows anyone to portray themselves as a world-class ballerina or a pop superstar like Bruno Mars. "With our framework, we create a variety of videos, enabling untrained amateurs to spin and twirl like ballerinas, perform martial arts kicks or dance as vibrantly as pop stars," the researchers stated in their paper. "Using pose detections as an intermediate representation between source and target, we learn a mapping from pose images to a target subject's appearance," the team explained. Using NVIDIA TITAN Xp and GeForce GTX 1080 Ti GPUs, with the cuDNN-accelerated PyTorch deep learning framework for both training and inference, the team first trained their conditional generative adversarial network on video of amateur dancers performing a range of poses filmed at 120 frames per second.

OneDrive and SharePoint are getting a bit smarter with search thanks to AI, with Microsoft unveiling automated transcription and image search. Microsoft has been slowly feeding artificial intelligence and machine learning tools into Office 365, and this latest round includes the ability to analyse audio files and other rich content stored in Microsoft 365, notably OneDrive and SharePoint, to make it easier to find video clips, audio recordings and images without scrolling endlessly. "A key to being productive is leveraging existing content so you're not reinventing the wheel," said Omar Shahine, partner director of program management for OneDrive and SharePoint, in a post. "Historically this has been challenging due to the exponential growth of digital content, particularly with image, video, and audio files. Until now, these rich file types have been cumbersome to manage and painful to sift through to find what you need, when you need it."