Computer Vision News - August 2021

MIDL Runner Up 22 He also plans to test whether images should be warped first before features are extracted, or vice versa. Warping is a non-linear image alignment , where all parts of the image are distorted. Winning the runner-up prize at a prestigious conference like MIDL is no mean feat for anyone, but for someone in the first year of their PhD , it is extra special. Why does Steffen think the judges picked his work above so many others? “I think what makes it stand out is that it is generally applicable,” he responds. “In the medical domain, a lot of research is very specific to a limited number of applications, whereas with this work I strive to be general. Also, I evaluated on three completely different biomedical datasets, which is very unusual for this community. People generally stick to one dataset and if it works that’s enough. By going the extra mile, I think this is a really important contribution.” The work was positively received by participants too, who were inspired by its novelty. “People really liked the idea of training these feature extractors specifically for the dataset, as opposed to what they do in other areas of computer vision where they’ll just use VGG-net,” he adds. Best of M I D L