Computer Vision News - May 2020

2 Summary Co puter Vision Challenge 32 on a previous answer. “In the 2019 challenge, we had median as a metric. Half of the images could be predicted really badly. This year, we won’t use median, we’ll use something less robust and very hard to beat. That is why participants should try to develop something stable . It should be stable first of all.” Finally, Simon declares that he is eager to see something that isn’t just a traditional simple convolutional neural network. Deep learning, but with a twist, perhaps? “We had several twists, but the base was just a simple convolutional neural network,” he says. “This may be something very interesting.”