Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
By Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, Hartwig Adam. 2018
17,165 citations recorded by the source corpus.
Spatial pyramid pooling module or encode-decoder structure are used in deep neural networks for semantic segmentation task. The former networks are able to encode multi-scale contextual information by probing the incoming features with filters or pooling operations at multiple rates and multiple effective fields-of-view, while the latter networks can capture sharper object boundaries by gradually recovering the spatial information. In this work, we propose to combine the advantages from both methods. Specifically, our proposed model, DeepLabv3+, extends DeepLabv3 by adding a simple yet effective decoder module to refine the segmentation results especially along object boundaries. We further explore the Xception model and apply the depthwise separable convolution to both Atrous Spatial Pyramid Pooling and decoder modules, resulting in a faster and stronger encoder-decoder network. We demonstrate the effectiveness of the proposed model on PASCAL VOC 2012 and Cityscapes datasets, achieving the test set performance of 89.0\% and 82.1\% without any post-processing. Our paper is accompanied with a publicly available reference implementation of the proposed models in Tensorflow at \url{this https URL}.
Subfields: Computer Vision and Pattern Recognition
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Ming-Hsuan Yang
- Xiaogang Wang
- Trevor Darrell
- Jitendra Malik
- Yi Yang
- Alan L. Yuille
Citations in the corpus
- Axial-DeepLab: Stand-Alone Axial-Attention for Panoptic Segmentation (2020, methodology): “05 Xception-71 [21,16] 24.”
- CCNet: Criss-Cross Attention for Semantic Segmentation (2018, neutral): “Deeplabv3+ [8] and DPC [3] both use a more stronger back-bone ( i.e ., Xception-65 & 71 vs. ResNet-101).”
- VOLO: Vision Outlooker for Visual Recognition (2021, neutral): “6 DeepLabv3+ [7] Xception-65 [9] ImgNet-1k 79.”
- Panoptic Feature Pyramid Networks (2019, methodology): “Our approach, which is a minimal extension to FPN, is able to achieve strong results compared to systems like DeepLabV3+ [12], which have undergone many design iterations.”
- Visual Prompt Tuning (2022, background): “Backbone ViT-L/16 ResNet-101 Method Full [83] Head Only Bias VPT-deep VPT+Bias Full [8] mIoU-SS 48.”
- Masked-attention Mask Transformer for Universal Image Segmentation (2021, background): “Although such specialized architectures [6, 10, 24, 37] have advanced each individual task, they lack the flexibility to generalize to the other tasks.”
- Per-Pixel Classification is Not All You Need for Semantic Segmentation (2021, methodology): “, ResNet [20]), MaskFormer outperforms DeepLabV3+ [8] by 1.”
- K-Net: Towards Unified Image Segmentation (2021, methodology): “In Table 3b, we further compare K-Net with other state-of-the-art methods [9, 69] with test-time augmentation on the validation set.”
- Reviving Iterative Training with Mask Guidance for Interactive Segmentation (2021, background): “We train the networks based on HRNet18s, HRNet18, and HRNet32 on 1, 2, and 4 Tesla P40, respectively, and use a single Tesla P40 to train the DeepLabV3+ model.”
- Occluded Video Instance Segmentation: A Benchmark (2021, background): “Our work is of course relevant to some image-level recognition tasks, such as semantic segmentation [40,8, 9], instance segmentation [17,21,28], panoptic segmentation [27,62,32], large vocabulary instance segmentation [16, 60], etc.”
- Simple Copy-Paste is a Strong Data Augmentation Method for Instance Segmentation (2020, neutral): “On both PASCAL detection and PASCAL semantic Model mIOU DeepLabv3+ † [4] 84.6 ExFuse † [70] 85.8 Eff-B7 [73] 85.2 Eff-L2 [73] 88.7 Eff-B7 NAS-FPN 83.9 w/ Copy-Paste pre-training (+2.7) 86.6 Table 6.”
- Getting to 99% Accuracy in Interactive Segmentation (2020, background): “[7] Xception [24] DeepLab-v3+ [25] 1:4 none DEXTR Maninis et al.”
- Iteratively Trained Interactive Segmentation (2018, methodology): “Adopting the state-of-the-art DeepLabV3+ architecture [8] for our network, we demonstrate that applying the iterative training procedure yields significantly improved results which surpass the state-of-the-art both for interactively creating segmentations from scratch and for correcting segmentation”
- MaX-DeepLab: End-to-End Panoptic Segmentation with Mask Transformers (2020, methodology): “3, our decoder is stacked L times, traversing output strides (4, 8, and 16 [16, 61]) multiple times.”
- Fast Neural Network Adaptation via Parameter Remapping and Architecture Search (2020, methodology): “For fair comparison with Auto-DeepLab (Liu et al., 2019a) which searches the back-bone architecture on DeepLabv3 and retrains the searched network on DeepLabv3+ (Chen et al., 2018b), we adapt the parameters of the target architecture Arch t to DeepLabv3+ framework.”
- EfficientDet: Scalable and Efficient Object Detection (2019, methodology): “35% - 298B DeepLabV3+ (Xception) [4] 80.”