DevDiary

I could check out another deep learning meetup. This time it was hosted at Google's Isar Valley here in Munich. The three interesting walks were about the following topics: Visual Sentiment Analysis with Deep Convolutional Neural Networks (by Dr. Damien Borth, DFKI) Strategies for AI Deployment (by Henrik Klagges, TNG Consulting GmbH) DGX-1 and SATURNV: The World’s Most Efficient Supercomputer for AI and Deep Learning (by Ralph Hinsche, Nvidia) In the third talk of Nvidia, we were also able to hold a test sample of the latest Tesla P100 in our hands, which is one of the building blocks of Nvidia's deep learning super computers called DGX-1 . This is a nice super toy that every AI-researcher would like to have under the Christmas tree. Unfortunately, a single device costs more than 100.000 US-Dollar.

I had the pleasure to check out Intel's first AI days in Europe. At ICM in Munich, Intel presented their latest advancements in Artificial Intelligence and Deep Learning in both hardware and software. As one of the biggest player in the hardware industry, they talked a lot about the next wave Xeon CPUs called Lake Crest , that is optimized for Deep Learning. Furthermore, a representative of Nervana Systems introduced their deep learning platform, which has been acquired by Intel for more than 400 Mio. US-Dollar in October 2016. Additionally, they talked a lot about low-level optimizations that they have done in order to accelerate many deep learning using Intel hardware, such as Intel Math Kernel Library ( MKL ). In some examples, they shows amazing improvements by a factor of up to 400. This sounds to good to be true in my ears, but even half of that is more than welcome! They presented their Neon framework , which feels to be in between TensorFlow and Keras, as we...

I finally finished my Master's Thesis in the Computer Vision chair at TUM. In the course of this thesis, I analyzed existing deep learning approaches to predict future frames in videos. Based on these findings and other modern deep learning practices, such as batch normalization, scheduled sampling to improve recurrent network training or ConvLSTMs, we were able to reach or event outperform state-of-the-art performance in future frame generation. So far, many people asked me about the practical application of frame prediction. Unfortunately, it won't tell us the end of any cliff-hanger movie such as Inception, but the main purpose of such a system is not to generate a perfect forecast of the long-term continuation of any movie clip. This completely impossible in my opinion, since there is not always a wrong or right in many situations. A neural network cannot be able to predict every decision made by all objects inside the scene. Furthermore, the pose of the camera or the ...