###################################################################
# #
# Deep Sketch Project #
# Xiong Duan (hgysdx@163.com) #
# #
###################################################################
Deepsketch
This code is incomplete corresponding to the paper, see more in http://dx.doi.org/10.1016/j.neucom.2016.04.046.
New version can be downloaded in https://github.com/XiongDuan/deepsketch.
If you use this demo in your project, we appreciate it if you cite an appropriate subset of our paper:
@article{wang2016deep, title={Deep Sketch Feature for Cross-domain Image Retrieval}, author={Wang, Xinggang and Duan, Xiong and Bai, Xiang}, journal={Neurocomputing}, year={2016}, publisher={Elsevier} }
The step of using deepsketch demo code.
-
dataset.
The skecth dataset can be downloaded in http://cybertron.cg.tu-berlin.de/eitz/projects/classifysketch/. And the corresponding image dataset can be downloaded in http://pan.baidu.com/s/1eSHfRdK
-
toolbox.
This project need download Piotr's Matlab Toolbox (https://pdollar.github.io/toolbox/), edgebox-toolbox(https://github.com/pdollar/edges) and caffe(https://github.com/BVLC/caffe) to support. For the windows-caffe, you can setup your environment by using(https://github.com/happynear/caffe-windows ).
-
put the deepsketch, Piotr's Matlab Toolbox and edgebox-toolbox in caffe-windows-master\matlab.
Then get in the dir deepsketch and edit deepsketch.m in your matlab wrapper and change all the paths in the code.
-
run deepsketch.m if there is no error.