Revision history [back]

Hi all,

While using the SURF FeatureDetector and DescriptorExtractor in Java, I discovered that the default parameters are the following: %YAML:1.0 name: "Feature2D.SURF" extended: 0 hessianThreshold: 100. nOctaveLayers: 3 nOctaves: 4 upright: 0 According to the documentation ( http://docs.opencv.org/modules/nonfree/doc/feature_detection.html ), the default parameters are the following: extended=1 nOctaveLayers=2 nOctaves=4 upright=0 and if I am not mistaken, usually the hessianThreshold is set to 300-500, pointed out in the same page of the documentation.

So, my first question is why there is that difference, and if that occurs to other features?

Secondly, the need of setting the default parameters, forced me to have to write a temp file, in order to give to the FeatureDetector and DescriptorExtractor the default parameters. I think that it is quite an overhead having to write an tempfile, and read the file, for both detector and extractor, so as to give parameters. So, my second question is, why there is not an easier way for passing parameters??

Thanks in advance.

Hi all,

While using the SURF FeatureDetector and DescriptorExtractor in Java, I discovered that the default parameters are the following: %YAML:1.0 name: "Feature2D.SURF" extended: 0 hessianThreshold: 100. nOctaveLayers: 3 nOctaves: 4 upright: 0 According to the documentation ( http://docs.opencv.org/modules/nonfree/doc/feature_detection.html ), the default parameters are the following: extended=1 nOctaveLayers=2 nOctaves=4 upright=0 upright=1 and if I am not mistaken, usually the hessianThreshold is set to 300-500, pointed out in the same page of the documentation.

So, my first question is why there is that difference, and if that occurs to other features?

Secondly, the need of setting the default parameters, forced me to have to write a temp file, in order to give to the FeatureDetector and DescriptorExtractor the default parameters. I think that it is quite an overhead having to write an tempfile, and read the file, for both detector and extractor, so as to give parameters. So, my second question is, why there is not an easier way for passing parameters??

Thanks in advance.

Hi all,

While using the SURF FeatureDetector and DescriptorExtractor in Java, I discovered that the default parameters are the following: %YAML:1.0 name: "Feature2D.SURF" extended: 0 hessianThreshold: 100. nOctaveLayers: 3 nOctaves: 4 upright: 0 According to the documentation ( http://docs.opencv.org/modules/nonfree/doc/feature_detection.html ), the default parameters are the following: extended=1 nOctaveLayers=2 nOctaves=4 upright=1 upright=0 and if I am not mistaken, usually the hessianThreshold is set to 300-500, pointed out in the same page of the documentation.

So, my first question is why there is that difference, and if that occurs to other features?

Secondly, the need of setting the default parameters, forced me to have to write a temp file, in order to give to the FeatureDetector and DescriptorExtractor the default parameters. I think that it is quite an overhead having to write an tempfile, and read the file, for both detector and extractor, so as to give parameters. So, my second question is, why there is not an easier way for passing parameters??

Thanks in advance.