In the review, written by Leon Gatys, Matthias Bethge and myself, we discuss recent advances in texture synthesis using Convolutional Neural Networks (CNNs) that were motivated by visual neuroscience and have led to a substantial advance in image synthesis and manipulation in computer vision. We also discuss how these advanecs can in turn inspire new research in visual perception and computational neuroscience.
Our psychophysical evaluation of our CNN-based texture model is now available on bioRxiv. In the study led by Tom Wallis, we compared our recent parameteric model of texture appearance (CNN model) that uses the features encoded by a deep convolutional neural network (VGG-19) with two other models: the venerable Portilla and Simoncelli model (PS) and an extension of the CNN model in which the power spectrum is additionally matched.
We just put a preprint on arXiv describing a number of improvements to the style transfer algorithm we developed a while ago. These new features include spatial control, color control and scale control.
Spatial control: applying different styles to different parts of the image (panel b).
Color control: transferring only the style of a painting, but keeping the colors of the original photograph (panel c). You can find additional examples in our blog post on blog.deepart.io.
Scale control: combine small-scale features of one style with large-scale features of another.
In a new paper that just came out in the Journal of Neuroscience we investigate how unobserved fluctuations in attentional state can induce correlated variability among neuronal populations. Interestingly, we found that an extremely simple model that treats attentional state as a shared gain can explain a wide variety of experimental findings on correlated variability.
Our paper A neural algorithm of artistic style (posted on arXiv on Aug 26) is ranked #9 among the most widely discussed and shared academic papers in 2015. See here for the top 100 list and the Altmetric report for our paper. Remarkably, in the all-time stats it ranks #102 out of more than 4.6 million articles.
We have a new paper published today in Science. Xiaolong Jiang in Andreas Tolias’ lab at Baylor College of Medicine recorded, labeled, and classified over 1600 neurons in mouse visual cortex and characterized their connectivity. Based on his remarkable work we discovered three simple connectivity rules that capture most of the structure of the connectivity matrix:
- There are two master regulators with distinct input profiles.
- Interneuron-selective interneurons are neither self-inhibitory nor locally recruited.
- Pyramidal-neuron-targeting interneurons are self-inhibitory and locally recruited.
We recently founded Pro-Test Germany, a non-profit organization by young scientists with the goal of communicating and educating about the need of using animals in research.
Whether or not we should use animals in research is a controversial question. While it is clear to a neuroscientist like myself that we need animal research to make progress in basic research and medicine, it may not be as obvious to someone who has never done science or even met a scientist.
Unfortunately, in Germany the public dispute on this topic has been dominated by animal rights activists with extreme views. Scientists, on the other hand, have mostly preferred to do their work instead of engaging with the public. After the recent events in Tübingen that forced Max Planck director Nikos Logothetis to give up his work with macaque monkeys, it has become clear to many of us that this needs to change. Scientists have to speak up, get out there, and talk about their work: What do we do? Why do we do it? Why is it necessary? How does it help society? To help answering these questions for everybody, we started Pro-Test Germany. We will have a website, be active on social media and have public actions in Tübingen. Visit our website or Facebook page and stay tuned!
I am a scientist and entrepreneur based in Tübingen, Germany. Working at the intersection of computational neuroscience, machine learning and computer vision, I would like to understand how neural systems – both biological and artificial – perform visual perception. I am also a co-founder of Layer7 AI, where I serve as Lead Machine Learning engineer, and DeepArt.io.