CCSE Researchers Score Big in International Challenge

Pascal VOC2012 challenge is one of the most prestigious Image Segmentation Challenge for Deep Learning and Artificial Intelligence tasks. It is the gold standard for Bench Marking visual object classification algorithms. This challenge is organized by the University of Oxford. The task is to make a Neural Network powerful enough to acquire as high a meanIOU score for the provided test data set by the competition where higher score correlates to identification accuracy.

Mr. Sharif Amit Kamran, a research associate at the Center for Cognitive Skill Enhancement at IUB under the guidance of Dr. Ali Shihab Sabbir, Director, CCSE, have developed a novel and efficient Deep Convolutional Neural Network for Image Segmentation task called FCN-2s_Dilated_VGG19 which scored 69.0 percent meanIOU score in the Segmentation Challenge beating many high scoring models from UC Berkeley,University of Oxford and others.


Model name : Dilated FCN using VGG19 and Skip Architectures
Short Form : FCN-2s_Dilated_VGG19