Method Details
Details for method 'DRN_CRL_Coarse'
Method overview
name | DRN_CRL_Coarse |
challenge | pixel-level semantic labeling |
details | Semantic image segmentation, which aims at assigning pixel-wise category, is one of challenging image understanding problems. Global context plays an important role on local pixel-wise category assignment. To make the best of global context, in this paper, we propose dense relation network (DRN) and context-restricted loss (CRL) to aggregate global and local information. DRN uses Recurrent Neural Network (RNN) with different skip lengths in spatial directions to get context-aware representations while CRL helps aggregate them to learn consistency. Compared with previous methods, our proposed method takes full advantage of hierarchical contextual representations to produce high-quality results. Extensive experiments demonstrate that our methods achieves significant state-of-the-art performances on Cityscapes and Pascal Context benchmarks, with mean-IoU of 82.8\% and 49.0\% respectively. |
publication | Dense Relation Network: Learning Consistent and Context-Aware Representation For Semantic Image Segmentation Yueqing Zhuang ICIP |
project page / code | https://github.com/zhuangyqin/DRN.git |
used Cityscapes data | fine annotations, coarse annotations |
used external data | ImageNet |
runtime | n/a |
subsampling | no |
submission date | February, 2018 |
previous submissions |
Average results
Metric | Value |
---|---|
IoU Classes | 82.6992 |
iIoU Classes | 61.0556 |
IoU Categories | 91.7857 |
iIoU Categories | 80.4749 |
Class results
Class | IoU | iIoU |
---|---|---|
road | 98.8277 | - |
sidewalk | 87.8252 | - |
building | 93.9967 | - |
wall | 65.4096 | - |
fence | 63.6886 | - |
pole | 70.1868 | - |
traffic light | 77.4079 | - |
traffic sign | 81.4412 | - |
vegetation | 93.8675 | - |
terrain | 73.4726 | - |
sky | 95.7971 | - |
person | 87.9066 | 70.9523 |
rider | 75.2043 | 53.3335 |
car | 96.4198 | 90.7995 |
truck | 76.7718 | 47.0926 |
bus | 91.5292 | 56.0492 |
train | 90.4161 | 55.9716 |
motorcycle | 72.2993 | 49.9733 |
bicycle | 78.8164 | 64.2731 |
Category results
Category | IoU | iIoU |
---|---|---|
flat | 98.7801 | - |
nature | 93.6028 | - |
object | 76.0845 | - |
sky | 95.7971 | - |
construction | 94.1553 | - |
human | 88.0284 | 71.95 |
vehicle | 96.0519 | 88.9999 |