Computer Vision News - February 2022

50 Presentation from WACV 2022 learning is the only way to analyze this in an efficient manner. ” It is clear when they speak that Izabela and Johannes are both highly passionate about this work. “ What excites me is that we are working with truly experimental imaging data, ” Johannes tells us. Izabela adds: “ I’m passionate because I can see the problems we work on have real applications. We can drive discoveries, Previously, biologists would be tasked with ma nually analyzing this kind of data and counting tumor cells . They did not have the capabilities to look at the whole specimen at the same time, so they would only look at small parts. Deep learning has been an enabler for unbiased analysis, making it possible to explore multiple whole animals or organs in rapid time. “ The data is so large, ” Johannes points out. “ We’re talking about single images of 10,000 by 10,000 by 10,000 pixels.Machine State of the art methods are not designed to handle multi-channel data with often inconsistent annotations, thus justifying the design of a novel architecture and training pipeline tailored to our problem. Vascular structure of a whole mouse brain imaged with light sheet microscopy. Source: Nature Methods. 3D reconstruction of Light Sheet Microscopy- imaged mouse. Source : Nature Neuroscience.