Traditional Inspired Network

This repository contains the implementation details of our paper:

"TRADITIONAL METHOD INSPIRED DEEP NEURAL NETWORK FOR EDGE DETECTION"
by Jan Kristanto Wibisono , Hsueh-Ming Hang

Dependencies

• Python 3.7
• Pytorch 1.4

Network Structure

Our systems contain three basic modules: Feature Extractor, Enrichment, and Summarizer, which roughly correspond to gradient, low pass filter, and pixel connection in the traditional edge detection schemes.

Evaluation

Comparison of complexity and accuracy performance among various edge detection schemes. Our proposed methods (Green). BDCN family (Red). Other methods (Blue). ODS (Transparent label). Number of Parameter (Orange label)

Todo:

Testing

    python inference.py
	

Manual Input

    python test.py

Training

    python train.py

Citing

Thanks for your interest in our work, please consider citing:

    @INPROCEEDINGS{9190982,
      author={J. K. {Wibisono} and H. -M. {Hang}},
      booktitle={2020 IEEE International Conference on Image Processing (ICIP)}, 
      title={Traditional Method Inspired Deep Neural Network For Edge Detection}, 
      year={2020},
      volume={},
      number={},
      pages={678-682},
    }