SAIN2018 Efficient Yet Deep Convolutional Neural Networks for Semantic Segmentation

Our paper has been accepted to 2018 International Symposium on Advanced Intelligent Informatics (SAIN).

IEEE Xplore Digital Library

https://ieeexplore.ieee.org/document/8673354

Arxiv Pre-print

https://arxiv.org/abs/1707.08254

Researchgate

https://www.researchgate.net/publication/318720999_Efficient_Yet_Deep_Convolutional_Neural_Networks_for_Semantic_Segmentation

Please cite our paper if you use our codes or material in your work:

Youtube Demo

Citation

@inproceedings{kamran2018efficient,
  title={Efficient yet deep convolutional neural networks for semantic segmentation},
  author={Kamran, Sharif Amit and Sabbir, Ali Shihab},
  booktitle={2018 International Symposium on Advanced Intelligent Informatics (SAIN)},
  pages={123--130},
  year={2018},
  organization={IEEE}
}

Score and Leaderboard

• FCN2s-Dilated-VGG16 Mean Iou score 67.6 percent FCN2s-Dilated-VGG16
• FCN2s-Dilated-VGG19 Mean Iou score 69 percent FCN2s-Dilated-VGG19

Installation

Make caffe with python wrapper. Detailed Instruction below

Models

• FCN2s-Dilated-VGG16 download link - FCN2s-Dilated-VGG16
• FCN2s-Dilated-VGG19 download link - FCN2s-Dilated-VGG19

This models were only trained on SBD and VOC data and for 21 classes segmentation task for PASCAL VOC2012 Segmentation Challenge.

Will be uploading the net trained on NYUDv2 dataset and Pascal-Context later on. Keep an eye on the page.

Demo

Open demo.py and change line 29 for running demo with different images. Run demo.py

Tutorial

A tutorial with elaborated instructions for running the inference is provided at ModelDepot.io

Surgery + Training on VOC2012 dataset

First read surgery-instructions.txt for details.

Then read training-instructions.txt for details.

Training on SBD dataset

To recreate our result you have to first train on VOC2012 dataset and then SBD dataset.

Read the training-sbd-instructions for details.

License

The code is released under the MIT License, you can read the license file included in the repository for details.

Caffe

Caffe is a deep learning framework made with expression, speed, and modularity in mind. It is developed by Berkeley AI Research (BAIR)/The Berkeley Vision and Learning Center (BVLC) and community contributors.

Check out the project site for all the details like

• DIY Deep Learning for Vision with Caffe
• Tutorial Documentation
• BAIR reference models and the community model zoo
• Installation instructions

and step-by-step examples.

Custom distributions

• Intel Caffe (Optimized for CPU and support for multi-node), in particular Xeon processors (HSW, BDW, Xeon Phi).
• OpenCL Caffe e.g. for AMD or Intel devices.
• Windows Caffe

Community

Please join the caffe-users group or gitter chat to ask questions and talk about methods and models. Framework development discussions and thorough bug reports are collected on Issues.

Happy brewing!

License and Citation

Caffe is released under the BSD 2-Clause license. The BAIR/BVLC reference models are released for unrestricted use.

Please cite Caffe in your publications if it helps your research:

@article{jia2014caffe,
  Author = {Jia, Yangqing and Shelhamer, Evan and Donahue, Jeff and Karayev, Sergey and Long, Jonathan and Girshick, Ross and Guadarrama, Sergio and Darrell, Trevor},
  Journal = {arXiv preprint arXiv:1408.5093},
  Title = {Caffe: Convolutional Architecture for Fast Feature Embedding},
  Year = {2014}
}