Spectral Residual Saliency map output does not match actual implementation

asked 2015-07-28 15:54:14 -0600

The implementation of Spectral Saliency algorithm in the opencv extra module (StaticSaliencySpectralResidual::computeSaliencyImpl) does not generate a saliency map similar to that of the implementation specified by Hou in [http://www.klab.caltech.edu/~xhou/pap...]. I have tried generating maps using the same sample image that was used in the paper and got different results.

Comparison - OpenCV vs. Hou implementation https://drive.google.com/file/d/0B9yd...

edit retag flag offensive close merge delete

add a comment

1 answer

Sort by » oldest newest most voted

answered 2015-07-29 02:59:28 -0600

StevenPuttemans

20029 ●16 ●82 ●207 http://stevenputtemans...

Hmm what do you expect? If you want the exact same output as the paper, then you should

Adapt the existing code to get a better end result and then push your changes to the repository
Use the original software from the author

Keep in mind that these algorithms are contributed by people who are eagerly in helping out and providing new functionality. Ofcourse you cannot expect a complete 1:1 match of results... on the other hand, could you show you are doing your visualisation? It seems to me that you are just rounding values for visualisation purposes and therefore do not get the same results.