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

Sig53 Dataset issue #22

Closed
Jie-su opened this issue Nov 30, 2022 · 4 comments
Closed

Sig53 Dataset issue #22

Jie-su opened this issue Nov 30, 2022 · 4 comments

Comments

@Jie-su
Copy link

Jie-su commented Nov 30, 2022

Hi,

I am trying to use Sig53 datasets for my research, but I met some problems with it.

  1. Does the sig53 equal to WidebandSig53? Since the impaired version of WindebandSig53 only consists of 250000 samples with a length of 262144. However, sig53 contains 5.2M samples in the paper.

  2. The WidebandSig53 dataset's target contains multi-labels (as a list), and there are no documents related to the transformation.

Thanks,

Jie,

@lboegner
Copy link
Contributor

Hi Jie-su,

Glad to hear you’re interested in Sig53 for your research!

  1. Sig53 and WidebandSig53 are two different datasets:

    • Sig53 contains shorter examples of 4,096 IQ samples with a single signal class in each. This dataset is primarily meant to be used for signal classification research. More details on this dataset can be found in this paper: https://arxiv.org/pdf/2207.09918.pdf

    • WidebandSig53, on the other hand, contains longer examples of 262,144 IQ samples with potentially multiple, non-overlapping (in time and frequency) signal classes present. This dataset is primarily meant to be used for signal detection & signal recognition research. More details on this dataset can be found in this paper: https://arxiv.org/pdf/2211.10335.pdf

  2. If using the SignalData object, the label for the WidebandSig53 dataset will be a list of SignalDescriptions, where each signal present in the data example has a SignalDescription element in the list. The WidebandSig53 dataset uses the DescToListTuple target transform at generation time, so when it is stored to disc, each example’s label is saved as a list of lists (rather than a list of SignalDescriptions), where the inner lists encompass the signal class, start, stop, center frequency, bandwidth, and SNR of each individual signal in the example.

Since I think the “issue” here is that we need better dataset documentation, I think we can close this as a duplicate of the existing issue @gvanhoy opened “Documentation: Better Dataset Documentation #18." Feel free to continue commenting with further questions in the meantime!

@Jie-su
Copy link
Author

Jie-su commented Dec 1, 2022

Hi,

Thanks for your explanation.

But I can't see how to generate Sig53 from the documentation, so does it already released or not?

Thanks,

Jie

@lboegner
Copy link
Contributor

lboegner commented Dec 1, 2022

Both datasets have been released. For examples and documentation on how to generate and use each dataset, check out the following notebooks in the examples directory:
- Sig53: https://github.com/TorchDSP/torchsig/blob/main/examples/00_example_sig53_dataset.ipynb
- WidebandSig53: https://github.com/TorchDSP/torchsig/blob/main/examples/03_example_widebandsig53_dataset.ipynb

@Jie-su
Copy link
Author

Jie-su commented Dec 2, 2022

Thanks!

Jie

@Jie-su Jie-su closed this as completed Dec 2, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants