Method Details

Details for method 'DeepLabv3+'


Method overview

name DeepLabv3+
challenge pixel-level semantic labeling
details 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 will provide more details in the coming update on the arXiv report.
publication Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation
Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, Hartwig Adam
project page / code
used Cityscapes data fine annotations, coarse annotations
used external data ImageNet
runtime n/a
subsampling no
submission date March, 2018
previous submissions


Average results

Metric Value
IoU Classes 82.1371
iIoU Classes 62.4329
IoU Categories 91.9972
iIoU Categories 81.9424


Class results

Class IoU iIoU
road 98.6939 -
sidewalk 87.0411 -
building 93.9102 -
wall 59.4754 -
fence 63.7375 -
pole 71.3946 -
traffic light 78.163 -
traffic sign 82.1568 -
vegetation 93.9698 -
terrain 73.0359 -
sky 95.8471 -
person 87.953 73.0552
rider 73.2603 53.728
car 96.4068 91.3836
truck 78.0207 47.0738
bus 90.9143 58.8446
train 83.9089 56.3355
motorcycle 73.8367 53.2403
bicycle 78.8796 65.8018


Category results

Category IoU iIoU
flat 98.7719 -
nature 93.6587 -
object 77.1163 -
sky 95.8471 -
construction 94.1685 -
human 88.3159 74.1221
vehicle 96.1023 89.7626



Download results as .csv file

Benchmark page