CVPR disputes again: IBM's Chinese draft papers were accused of copying idea, who won the second place in the competition

A wave is not flat , Wave after wave ,CVPR 2022 It was also exposed that the thesis was plagiarized ！

These two days , The research team from China found that , Last year I participated in ICDAR The contest idea, I was IBM Take it and hit this year's CVPR.

And the most ironic thing about this is ,ICDAR 2021 It is from IBM Self undertaken .

at present , The author has sent the relevant evidentiary information to CVPR Of Program Chairs.

The heat of this matter is still fermenting .Reddit The heat of discussion is about to break through a thousand .

below , Let's use the first person , Restore the original author's readme .

「 My name is Xianbiao Qi, We have been doing research in the field of computer vision for more than ten years . I am writing this blog to expose a blatant act of plagiarism .IBM The Zurich Institute plagiarized our research .」

「 They didn't copy the text , They copied what we worked hard to make idea.」

Xianbiao Qi The thesis is 「PingAn-VCGroup's Solution for ICDAR 2021 Competition on Scientific Literature Parsing Task B: Table Recognition to HTML」, On 2021 year 6 Upload to arXiv, Partners have Jiaquan Ye, Xianbiao Qi and Yelin He wait forsomeone .

Besides , The code is also open source .

Address of thesis ：https://arxiv.org/abs/2105.01848

And be Xianbiao Qi The plagiarism team is from IBM Zurich Institute Ahmed Nassar, Nikolaos Livathinos, Maksym Lysak and Peter Staar wait forsomeone , The article is called 「TableFormer: Table Structure Understanding with Transformers」, Please polish your eyes . Address of thesis ：https://arxiv.org/abs/2203.01017

Xianbiao Qi Express ,Nassar And others' articles put our overall approach , And pretreatment and post-processing 、 Visualization part 、 The reasoning part 、 The system solutions given are copied in the past . There are even code and pre training models .

stay Reddit On , Many netizens expressed their support for the author's rights protection , At the same time, it also reminds him not to be emotional when giving evidence , It's hard to do .

Some netizens also pointed out , The evidence now listed is far from conclusive ：

The two teams are on the same issue , Use the same public data , The solutions are similar . But this does not mean that one party copied the other .

Last ,Xianbiao Qi The most profound lesson learned this time is ：「 Even open source code , Don't open source your own trained models .」

Read the details