Luc Kusters

Sorry if necrobumping but any update on this? This feature would greatly streamline my build process.

@peduajo have a look at my repo https://github.com/ljbkusters/YOLOv3-1D. It implements the network to work with 1D convolutions but the main structure remains the same. Specifically regarding the issue I have...

> I finally found some time to work on the PR again. I added an additional test following the suggestion by @DietBru, which tests for the marginals. While implementing that,...

Yes, I developed it myself, based on the [paper](https://www.researchgate.net/publication/257291758_The_Wigner_distribution-A_tool_for_time-frequency_signal_analysis-Part_II_Discrete_time_signals) I link in the docstring. Specifically equations 2.10 and 2.11. The implementation is just some manipulation of indices to get the...

> That sounds promising. If it is okay for you I'll try to modify the code in the PR with your algorithm, keeping the interface, when I find some time...