DCNAS: Densely Connected Neural Architecture Search for Semantic Image Segmentation

26 Mar 2020Xiong ZhangHongmin XuHong MoJianchao TanCheng YangWenqi Ren

Neural Architecture Search (NAS) has shown great potentials in automatically designing scalable network architectures for dense image predictions. However, existing NAS algorithms usually compromise on restricted search space and search on proxy task to meet the achievable computational demands... (read more)

PDF Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
RESULT LEADERBOARD
Semantic Segmentation ADE20K val DCNAS mIoU 47.12% # 3
Semantic Segmentation Cityscapes test DCNAS Mean IoU (class) 83.6% # 5
Semantic Segmentation PASCAL Context DCNAS mIoU 55.6 # 5
Semantic Segmentation PASCAL VOC 2012 test DCNAS Mean IoU 86.9% # 4