Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Generate all grayscale single-index spectrograms #199

Closed
atruskie opened this issue Oct 19, 2018 · 3 comments
Closed

Generate all grayscale single-index spectrograms #199

atruskie opened this issue Oct 19, 2018 · 3 comments
Assignees

Comments

@atruskie
Copy link
Member

Is your feature request related to a problem? Please describe.

The spectral indices are rendered as grayscale spectrogram images for the six spectral indices used in the false colour spectrogram colour-maps. Here an example grayscale spectral ACI spectrogram:

fi_location_sm304246_mcleodroad_20180912__aci

Describe the solution you'd like

Two fold:

  • By default generate the grayscale spectrograms for all spectral indices
  • Add a new option to the Towsey.Acoustic.yml config to limit or disable index image generation
    • The config option will have a name like GenerateIndicesImages which will take a value of All, CompositeOnly, SingularOnly, None

Describe alternatives you've considered

We have considered removing image generation all together from the Towsey.Acoustic analysis since most people want concatenated images produced fror multiple files - there's no use generating the images twice.

Additional context

The single-index spectrograms are very useful for examining the effects of different indices without the colour bias or confounding effects that are found in the false colour spectrograms.

@towsey
Copy link
Contributor

towsey commented Oct 22, 2018

The request is to generate single index (grey scale) concatenated 24-hour spectrograms of all the indices for which matrices of values are available. This will assist the search for those indices that detect acoustic events of interest. And this, in turn, is useful when constructing feature sets for ML purposes.
I agree that there is no need to construct any of the "partial" spectrogram images" where the ultimate intention is to produce 24 hour spectrograms.

@atruskie
Copy link
Member Author

atruskie commented Oct 22, 2018

@towsey the 24-hour clarification is not needed. The concatenateIndexFiles command (which produces 24-hour FCS) should have identical behaviour to the Towsey. Acoustic analysis. This modification as such is really a change to the data contract of an indices analysis and thus both bits of code need new behaviour.

Keep in mind we have no limitation on the length of a recording... Just the other day we were processing 24-hour files.

As to to the 'partial' images: images associated with the recording are the whole image. Concatenated images are super images.

towsey added a commit that referenced this issue Dec 18, 2018
Changes required in order deal with side effects of the pervious commit.
All this is in part resolution of Issue #199.
Yet TODO: Still need to check that Concatenation methods concatenate all the available grey scale spectrograms.
atruskie added a commit that referenced this issue Dec 19, 2018
The changes in ae5f673 broke many tests.

This commit fixes some but not all tests. More work needed.

Work for #199
@towsey towsey reopened this Jan 29, 2019
@towsey
Copy link
Contributor

towsey commented Jan 29, 2019

Reopened because these code written for this patch only generated the extra images in the concatenate path. More work required to do this for also the standard index generation path.

towsey added a commit that referenced this issue Jan 29, 2019
All single index greyscale spectrograms are now drawn. I changed the config file to get better grey scale rendering. Changed drawing of Hertz label so does not cover so much spectrogram real-estate.
atruskie pushed a commit that referenced this issue Jan 31, 2019
Removed code that LDSpectrogramRGB.cs that combined multiple grey scale spectrograms into one image.

More work for #199 and addressing feedback from #211
atruskie added a commit that referenced this issue Jan 31, 2019
Two of the failing tests were just due to different numbers of files being produced.

The other failed in an aritficial data scenario where the min and max of the matrix were equal. The resulting range (0) caused NaNs  when normalized. This bug was discovered now due to the dual auto-normalizations changes to the index properties config for R3D.

Fix is to revert to pre-defined bounds is auto-noramlization range is 0.

This fixes work done in 6ec16c6 for #199
atruskie added a commit that referenced this issue Feb 4, 2019
Adds rendering of ALL grayscale LD spectrograms to standard indices generation. Closes #199 (again)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants