DMNet-paddlepaddle
1.Introduction
Paddle implementation of Dynamic Multi-scale filters for Semantic Segmentation. Our team is computer vision(计算机幻觉).
Official Repository | mmsegmentation version | Paper
2.Results and models
2.1Cityscapes
The following results are from mmsegmentation version. Pretrained weighs can be downloaded from
R-101-D8 or baidu cloud with code hhnn.
| Method | Backbone | Crop Size | Lr schd | Inf time (fps) | mIoU | config | download |
|---|---|---|---|---|---|---|---|
| DMNet | R-101-D8 | 512×1024 | 80000 | – | 79.64 | config | model | log |
We only trained the Cityscapes dataset with backnone R-101-D8,the results:
| Method | Backbone | Crop Size | Lr schd | Inf time (fps) | mIoU | config | download |
|---|---|---|---|---|---|---|---|
| DMNet | R-101-D8 | 512×1024 | 80000 | – |
3.Quick start
3.1Prerequisites
- Linux
- Python 3.6+
- Paddlepaddle
- CUDA 10.0+
- GCC 5+
3.2Installation
a.Install dependencies
pip install -r requestments.txt
License
This project is released under the Apache 2.0 license.
Contributing
@luyuxuan
Citation
@InProceedings{He_2019_ICCV,
author = {He, Junjun and Deng, Zhongying and Qiao, Yu},
title = {Dynamic Multi-Scale Filters for Semantic Segmentation},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
month = {October},
year = {2019}
}
