11.03.2018       Выпуск 220 (05.03.2018 - 11.03.2018)       Интересные проекты, инструменты, библиотеки

TFSegmentation - Real-time Semantic Segmentation Comparative Study

Читать>>



Экспериментальная функция:

Ниже вы видите текст статьи по ссылке. По нему можно быстро понять ссылка достойна прочтения или нет

Просим обратить внимание, что текст по ссылке и здесь может не совпадать.

Real-time Semantic Segmentation Comparative Study

By: Mennatullah Siam, Mostafa Gamal, Moemen AbdelRazek, Senthil Yogamani, Martin Jagersand

The repository contains the official TensorFlow code used in the our paper RTSEG: REAL-TIME SEMANTIC SEGMENTATION COMPARATIVE STUDY for comparing different realtime semantic segmentation architectures.

Description

Semantic segmentation benefits robotics related applications especially autonomous driving. Most of the research on semantic segmentation is only on increasing the accuracy of segmentation models with little attention to computationally efficient solutions. The few work conducted in this direction does not provide principled methods to evaluate the different design choices for segmentation. In this paper, we address this gap by presenting a real-time semantic segmentation benchmarking framework with a decoupled design for feature extraction and decoding methods. The code and the experimental results are presented on the CityScapes dataset for urban scenes.

Models

NOTE: The pretrained weights for all the implemented models will be released soon. Stay in touch for the updates.

Reported Results

Test Set

ModelGFLOPs**Class IoUClass iIoUCategory IoUCategory iIoU
SegNet286.0356.134.279.866.4
ENet3.8358.324.480.464.0
DeepLab-70.442.686.467.7
SkipNet-VGG16-65.341.785.770.1
SkipNet-ShuffleNet2.058.332.480.262.2
SkipNet-MobileNet6.261.535.282.063.0

Validation Set

EncoderDecoderCoarsemIoU
MobileNetSkipNetNo61.3
ShuffleNetSkipNetNo55.5
ResNet-18UNetNo57.9
MobileNetUNetNo61.0
ShuffleNetUNetNo57.0
MobileNetDilationNo57.8
ShuffleNetDilationNo53.9
MobileNetSkipNetYes62.4
ShuffleNetSkipNetYes59.3

** GFLOPs is computed on image resolution 360x640. However, the mIOU(s) are computed on the official image resolution required by CityScapes evaluation script 1024x2048.

Usage

Run

The file named run.sh provide a good example for running different architectures. Have a look at this file.

Example to the running command written in the file:

python3 main.py --load_config=[config_file_name].yaml [train/test] [Trainer Class Name] [Model Class Name]

Main Dependencies

Python 3 and above
tensorflow 1.3.0/1.4.0
numpy 1.13.1
tqdm 4.15.0
matplotlib 2.0.2
pillow 4.2.1
PyYAML 3.12

All Dependencies

pip install -r [requirements_gpu.txt] or [requirements.txt]  

Citation

If you find RTSeg useful in your research, please consider citing our work:

@ARTICLE{2018arXiv180302758S,   
   author = {{Siam}, M. and {Gamal}, M. and {Abdel-Razek}, M. and {Yogamani}, S. and    
	{Jagersand}, M.},   
    title = "{RTSeg: Real-time Semantic Segmentation Comparative Study}",   
  journal = {ArXiv e-prints},   
archivePrefix = "arXiv",   
   eprint = {1803.02758},   
 primaryClass = "cs.CV",   
 keywords = {Computer Science - Computer Vision and Pattern Recognition},   
     year = 2018,   
    month = mar,   
   adsurl = {http://adsabs.harvard.edu/abs/2018arXiv180302758S},   
  adsnote = {Provided by the SAO/NASA Astrophysics Data System}   
}

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.



Лучшая Python рассылка



Разместим вашу рекламу

Пиши: mail@pythondigest.ru

Нашли опечатку?

Выделите фрагмент и отправьте нажатием Ctrl+Enter.

Система Orphus