Postdoc position at the Division of robotics, perception and learning at KTH Royal Institute of Technology.
Deep networks have been repeatedly proven to benefit from large amount of labelled training data. Collecting high-quality labelled data, at Google scale, is prohibitively laborious and expensive for most societal bodies and industrial sectors. In this project, the postdoc candidate will work on unifying methods for undersupervised learning which includes semi-supervised, weakly-supervised, and noisily-supervised learning techniques. The postdoc will be hired at Hossein Azizpour’s group at the division of robotics, perception and learning and the project is in collaboration with a twinning postdoc at Aalto university in the group led by Juho Kannala. The postdoc position is funded by WASP and is anticipated to involve two concrete applications in medical image analysis and autonomous driving.