BinaRep plagiarizes Temporal Binary Representation

[fedebecat]

Hi,

I wanted to prepare a pull request for integrating Temporal Binary Representation in Tonic, when I discovered BinaRep among your event encoding methods: https://tonic.readthedocs.io/en/latest/auto_examples/representations/plot_tobinarep.html#sphx-glr-auto-examples-representations-plot-tobinarep-py

Temporal Binary Representation is an encoding strategy that I published in 2020 in the proceedings of the 25th International Conference on Pattern Recognition (ICPR): https://www.computer.org/csdl/proceedings-article/icpr/2021/09412991/1tmhW5YGUSY
A preprint of the paper was also made available in October 2020: https://arxiv.org/pdf/2010.08946.pdf

The main idea of the paper was to collect binary frames with an arbitrarily small accumulation time and then merge N consecutive binary frames into a single frame just by performing a binary-to-decimal conversion. In this way, we convert a sequence of binary values for a certain pixel into a single decimal integer describing its temporal evolution.

The BinaRep paper, was published in 2022 at the International Conference on Image Processing, that is two years after TBR was published:
https://ieeexplore.ieee.org/document/9898061
The thing is that BinaRep proposed exactly the same approach as TBR.

The similarity can be easily seen by comparing a few lines of code in our implementation and the one integrated here in Tonic.

Original implementation of our approach:
https://github.com/fedebecat/tbr-event-object-detection/blob/master/src/tbe.py#L29

Implementation of BinaRep in Tonic:

tonic/tonic/functional/to_bina_rep.py

Line 61 in debdd2e

return np.sum(mask * frames, 0) / (2 ** mask.shape[0] - 1)

I have already informed IEEE, which hosts the proceedings of both papers. I would therefore ask you to remove references to the plagiarizing paper from the Tonic website or at least reference the correct paper (TBR) for the method instead of BinaRep.

Thank you for your understanding.

Federico

[fabrizio-ottati]

Hi! Sorry to hear that but, as you can imagine, we do not check plagiarism in papers 🤣 The code was added by the first author of the paper, as far as I remember.

Could you prepare a pull request in which you replace also the correct citation? I think, however, that we would need to wait until IEEE officially recognizes the plagiarism, before accepting the PR. What do you think, @biphasic? It seems to me that we could accept the PR ASAP, 'cause I see the similarity, but I wanted to know your opinion.

Needless to say, we just look at the code, so we would like to not be involved in the papers dispute :)

[fabrizio-ottati]

On a second thought, this matter is too sensitive for us to decide. Either get in touch with the other paper's author and reach an agreement, or we wait for IEEE to make it official. I am sorry, but doing otherwise would be irresponsible on our side.

[biphasic]

@fedebecat I'm happy to change the reference to the original paper, could you please open a PR for that?

[fedebecat]

Thank you both for acknowledging this issue.
Thanks @biphasic! How should I handle this? I am preparing a PR with the implementation of TBR (obviously it is very close to the one of binarep, as they do the same thing, but the implementation of TBR does not require to know the number of frames in advance and from what I can see it is quite faster). Should I add a completely new transformation called TBR with the appropriate reference? What about binarep, should I remove it in the PR or modify it in any way?

Thank you

[Barchid]

Hello @fedebecat @biphasic and @fabrizio-ottati ,

I am the first author of BinaRep (which I implemented in Tonic). Thanks to @biphasic for reaching me out about this issue.

The first thing I should write is a huge apology to @fedebecat and colleagues for this problem. To my greatest shame, we were not aware of your previous contribution. It seems to me pretty obvious that, apart from the use of several consecutive frames, BinaRep was not as novel as we thought back then.

Of course, I am perfectly fine for removing BinaRep from Tonic and replacing it with TBR.

Again, I am sorry for the inconvenience that @fedebecat or the other authors of TBR may have experienced because of this error. If you do not mind, I will get in touch with you to clean up the mess, notably regarding ours published papers.

Thank you all for your patience.

[biphasic]

Thanks for chiming in @Barchid . Then @fedebecat if you could change the implementation, the name and the reference in a PR that would be great. For the moment we'll remove BinaRep until @Barchid shows that it is sufficiently different

[fedebecat]

Thank you @Barchid for your message. Ok @biphasic, I will make the PR as soon as possible, thank you again!