The bounding-box propagating tool is an effective tool for generating bounding-box in image set by users' input.
Before you using this tool, you have to setup the envrioment below:
MATLAB R2016a (or later version)
openCV 2.4 (or later version)
Microsoft Visual Studio 2010 (or later version)
Then, you need to setup the RP method in MATLAB by using commands below:
**In MATLAB **
>>> addpath('compile')
>>> addpath('cmex')
>>> gbvs_install
>>> gbvs_compile
Next, you need to compile the file of MBS(?) saliency map by:
**In MATLAB **
>>> addpath('mex')
>>> compile_win
The GUI of this tool:
![GUI](https://github.com/zwm0426/bboxTool/blob/master/GUI_snap/GUI.JPG?raw=true)- You should select the input directory, output directory and evaluation directory (optional).
- Then, click "Run" button.
- After you click "Run" button, all the directory are set and cannot be changed.
Then, the program will choose the images from input image set.
![GUI](https://github.com/zwm0426/bboxTool/blob/master/GUI_snap/gui3.JPG?raw=true)- You can use your mouse dragging on the image to bound the object, those object would be used as the
- If the image is the noise image, you can just click the image and our algorithm will ignore this image.
After your annotation, this tool will proceed to propagate bounding-box. We would show every result here for you to see the result.
![GUI](https://github.com/zwm0426/bboxTool/blob/master/GUI_snap/gui4.JPG?raw=true)Optinally, if you choose to evaluate the quality of the result by Jacarrd socre, then the GUI would proceed to evaluation part.
![GUI](https://github.com/zwm0426/bboxTool/blob/master/GUI_snap/gui5.JPG?raw=true) ![GUI](https://github.com/zwm0426/bboxTool/blob/master/GUI_snap/gui6.JPG?raw=true)After all the procedure, this tool would show the image like this.
![GUI](https://github.com/zwm0426/bboxTool/blob/master/GUI_snap/gui7.JPG?raw=true)The whole procedure are done now!
- You could check result file in the output directory you choosed.
File: "ANNO_xxx.jpg" is the bounded by yourself.
File: "RESU_xxx.jpg" is the image we propageted your bounding-box.
- And you could also check the folder "EVALUATION", if you choosed with the ground-truth folder.
Thank you for using our tool.
##Credits
[1] S. Manen, M. Guillaumin, and L. Van Gool. Prime Object Proposals with Randomized Prim's Algorithm. In ICCV, 2013.
[2] P. F. Felzenszwalb and D. P. Huttenlocher Efficient graph-based image segmentation. In IJCV, 2004 http://people.cs.uchicago.edu/~pff/segment/
[3] Zhang, Jianming, et al. "Minimum barrier salient object detection at 80 fps." Proceedings of the IEEE International Conference on Computer Vision. 2015.