papers AI Learner
The Github is limit! Click to go to the new site.

Label Super Resolution with Inter-Instance Loss

2019-04-09
Maozheng Zhao, Le Hou, Han Le, Dimitris Samaras, Nebojsa Jojic, Danielle Fassler, Tahsin Kurc, Rajarsi Gupta, Kolya Malkin, Shahira Abousamra, Shroyer Kenneth, Joel Saltz

Abstract

For the task of semantic segmentation, high-resolution (pixel-level) ground truth is very expensive to collect, especially for high resolution images such as gigapixel pathology images. On the other hand, collecting low resolution labels (labels for a block of pixels) for these high resolution images is much more cost efficient. Conventional methods trained on these low-resolution labels are only capable of giving low-resolution predictions. The existing state-of-the-art label super resolution (LSR) method is capable of predicting high resolution labels, using only low-resolution supervision, given the joint distribution between low resolution and high resolution labels. However, it does not consider the inter-instance variance which is crucial in the ideal mathematical formulation. In this work, we propose a novel loss function modeling the inter-instance variance. We test our method on two real world applications: cell detection in multiplex immunohistochemistry (IHC) images, and infiltrating breast cancer region segmentation in histopathology slides. Experimental results show the effectiveness of our method.

Abstract (translated by Google)
URL

http://arxiv.org/abs/1904.04429

PDF

http://arxiv.org/pdf/1904.04429


Similar Posts

Comments