%0 Generic %1 liu2019autodeeplab %A Liu, Chenxi %A Chen, Liang-Chieh %A Schroff, Florian %A Adam, Hartwig %A Hua, Wei %A Yuille, Alan %A Fei-Fei, Li %D 2019 %K %T Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation %U http://arxiv.org/abs/1901.02985 %X Recently, Neural Architecture Search (NAS) has successfully identified neural network architectures that exceed human designed ones on large-scale image classification. In this paper, we study NAS for semantic image segmentation. Existing works often focus on searching the repeatable cell structure, while hand-designing the outer network structure that controls the spatial resolution changes. This choice simplifies the search space, but becomes increasingly problematic for dense image prediction which exhibits a lot more network level architectural variations. Therefore, we propose to search the network level structure in addition to the cell level structure, which forms a hierarchical architecture search space. We present a network level search space that includes many popular designs, and develop a formulation that allows efficient gradient-based architecture search (3 P100 GPU days on Cityscapes images). We demonstrate the effectiveness of the proposed method on the challenging Cityscapes, PASCAL VOC 2012, and ADE20K datasets. Auto-DeepLab, our architecture searched specifically for semantic image segmentation, attains state-of-the-art performance without any ImageNet pretraining.

@misc{liu2019autodeeplab, abstract = {Recently, Neural Architecture Search (NAS) has successfully identified neural network architectures that exceed human designed ones on large-scale image classification. In this paper, we study NAS for semantic image segmentation. Existing works often focus on searching the repeatable cell structure, while hand-designing the outer network structure that controls the spatial resolution changes. This choice simplifies the search space, but becomes increasingly problematic for dense image prediction which exhibits a lot more network level architectural variations. Therefore, we propose to search the network level structure in addition to the cell level structure, which forms a hierarchical architecture search space. We present a network level search space that includes many popular designs, and develop a formulation that allows efficient gradient-based architecture search (3 P100 GPU days on Cityscapes images). We demonstrate the effectiveness of the proposed method on the challenging Cityscapes, PASCAL VOC 2012, and ADE20K datasets. Auto-DeepLab, our architecture searched specifically for semantic image segmentation, attains state-of-the-art performance without any ImageNet pretraining.}, added-at = {2021-01-20T12:56:43.000+0100}, author = {Liu, Chenxi and Chen, Liang-Chieh and Schroff, Florian and Adam, Hartwig and Hua, Wei and Yuille, Alan and Fei-Fei, Li}, biburl = {https://www.bibsonomy.org/bibtex/279db680008ef264cb0141121152bcd53/philipphaas}, interhash = {12032b2793bab10d8199c64da666cd3c}, intrahash = {79db680008ef264cb0141121152bcd53}, keywords = {}, note = {cite arxiv:1901.02985Comment: To appear in CVPR 2019 as oral. Code for Auto-DeepLab released at https://github.com/tensorflow/models/tree/master/research/deeplab}, timestamp = {2021-01-20T12:56:43.000+0100}, title = {Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation}, url = {http://arxiv.org/abs/1901.02985}, year = 2019 }