MIL-CAM ACCEPTED TO ECCV 2020 WORKSHOP ON COMPUTER VISION PROBLEMS IN PLANT PHENOTYPING!

Diagram of the MIL-CAM model for minirhizotron root segmentation.

Congratulations to our labmates and collaborators: Guohao Yu, Alina Zare, Weihuang Xu, Roser Matamala, Joel Reyes-Cabrera, Felix B. Fritschi and Thomas E. Juenger!  Their paper, “Weakly Supervised Minirhizotron Image Segmentation with MIL-CAM” was recently accepted to the 16th European Conference on Computer Vision (ECCV) Workshop on Computer Vision Problems in Plant Phenotyping (CVPPP 2020).  

Their paper explores weakly supervised minirhizotron image segmentation as way to ease tedious labeling tasks for root segmentation.  The authors introduce a multiple instance class activation map (MIL-CAM) which learns from image-level labels to perform pixel-level segmentation.  Results show that the proposed method outperforms alternative methods for localization of root objects in minrhizotron imagery.

Guohao will present for the virtual conference on August 28th.  Check out the pre-print here!