Another CVPR 2022 paper was accused of plagiarism, and Ping An insurance researchers sued IBM Zurich team

Pay attention to the parallel stars

Never get lost

Institute of computer vision

official account ID｜ComputerVisionGzq

Study Group ｜ Scan the code to get the join mode on the homepage

Computer Vision Institute column

author ：Edison_G

「 They are not copying the text , It's plagiarism .」

from 《 Almost Human 》

01

summary

yesterday , Someone is here. reddit Post posting ,IBM Zurich Research Center plagiarized its own paper , And it was CVPR 2022 To receive the .

The poster wrote ：

My name is Xianbiao Qi, Is a computer vision researcher with more than ten years of research experience . I am writing this blog to complain about a serious case ：IBM Employees of the Zurich Research Center deliberately plagiarized our papers . They are not copying the text , It's plagiarism .

Qi The suspected plagiarized paper mentioned is entitled 「PingAn-VCGroup's Solution for ICDAR 2021 Competition on Scientific Literature Parsing Task B: Table Recognition to HTML（ Safety VCGroup by ICDAR 2021 Scientific literature analysis competition task B Solutions provided ：HTML Form identification ）」, The author is from Ping An insurance and Ping An technology , Upload arXiv The time is 2021 year 5 month ,2021 year 7 The code was released in January ,2021 year 9 Month issued Tablemaster Pre training model ,2021 year 11 Month issued tablemaster_mmocr docker Environmental Science . The competition mentioned in the paper is 2021 year 4 At the end of the month , from IBM Another team hosted .

Thesis link ：https://arxiv.org/pdf/2105.01848.pdf

As stated in the title of the paper , This article is for ICDAR 2021 Scientific literature analysis competition task B（HTML Form identification ） Provides a set of solutions . say concretely , They divided table content recognition into four subtasks ： Table structure identification 、 Text line detection 、 Text line recognition and box assignment . The table structure recognition algorithm is based on a robust image text recognition algorithm MASTER custom .PSENet Used to detect each text line in the table image . The text line recognition model is also built on MASTER above . Last , In the box allocation stage , They will PSENet The detected text box is associated with a structure item reconstructed by table structure prediction , And the recognizable content of the text line is filled in the corresponding item .

Accused of plagiarism IBM The paper is entitled 「TableFormer: Table Structure Understanding with Transformers（TableFormer： use Transformer Understanding table structure ）」, Upload arXiv The time is 2022 year 3 month , Submit to CVPR The time of is unknown .

Thesis link ：https://arxiv.org/pdf/2203.01017.pdf

In this paper ,IBM A new table structure recognition model is proposed by the researchers of , It improves the latest end-to-end deep learning model from two aspects （ From PubTabNet Of encoder-dual-decoder）. First , They introduced a new object detection decoder for table cells . In this way , They can go directly from PDF Get programmatic... From the source PDF Table cell contents of , This avoids customization OCR Decoder training . This architectural change has led to more accurate table content extraction . secondly , They will LSTM The decoder is replaced by based on transformer The decoder of .

Qi Express ,IBM Their methodology was plagiarized in this paper 、 Preprocessing 、 post-processing 、 Visual effect 、 Reasoning 、 System level solutions 、 Code 、 Pre training model, etc , but 「 In order not to be caught 」, They don't quote Qi Papers of others .

Next ,Qi A dozen pictures show what he thinks IBM Where there is plagiarism .

Qi The suspected plagiarism points displayed 1： methodology . He thinks that ,IBM Of the researchers copied their ideas , Rewrite the 「 contribution 」 part , And redrawn the structure diagram according to their diagram .

Qi The suspected plagiarism points displayed 2： Preprocessing .

Qi The suspected plagiarism points displayed 3： post-processing .

Qi The suspected plagiarism points displayed 4： Reasoning acceleration method

Qi The suspected plagiarism points displayed 5： skill .

Qi The suspected plagiarism points displayed 6： Text line detection and text line recognition .

Qi The suspected plagiarism points displayed 7： System level solutions .

Qi The suspected plagiarism points displayed 8： Visual effect .

Qi The suspected plagiarism points displayed 9： Misleading readers .

I'm looking Qi After the example given , Some people think these are very strong evidence ,IBM It is difficult for the team to prove that their papers are original ,Qi We should go to CVPR Procedure Committee complaints .

however , Some people think that , The evidence is not strong enough ：「 The two groups are working on the same problem , Using the same public data , Their solutions also have similarities . But this does not mean that one side copies the other . for example 「 Suspected of plagiarism 1」 The content shown in is not so close 」.

at present , We haven't seen it at the bottom of the post yet IBM The author's response .

however , It is worth mentioning that , This is not this year CVPR The only suspected plagiarism incident exposed . A few days ago , A netizen in YouTube It was revealed that , A paper from Seoul University copied a dozen other papers , At present, the author of this paper has apologized and withdrawn his manuscript .

Maybe , The top meeting papers need more effort in checking the duplicate .

Reference link ：https://www.reddit.com/r/MachineLearning/comments/vlpnuw/d_ibm_zurich_research_plagiarised_our_paper_and/

 THE END 

Please contact the official account for authorization.

The learning group of computer vision research institute is waiting for you to join ！

ABOUT

Institute of computer vision

The Institute of computer vision is mainly involved in the field of deep learning , Mainly devoted to face detection 、 Face recognition , Multi target detection 、 Target tracking 、 Image segmentation and other research directions . The Research Institute will continue to share the latest paper algorithm new framework , The difference of our reform this time is , We need to focus on ” Research “. After that, we will share the practice process for the corresponding fields , Let us really experience the real scene of getting rid of the theory , Develop the habit of hands-on programming and brain thinking ！

VX：2311123606

Previous recommendation  

• AI Help social security , The latest video abnormal behavior detection method framework

• Explode ！YOLOv6 The fast and accurate target detection framework is open source （ With source code download ）
  

• At present, the target detection model with the highest accuracy, the fastest efficiency and the smallest storage （ Download the source code ）

• Artificial intelligence AI Challenge the college entrance examination papers , As a result, the human race won this time ！
  

• CVPR Small target detection ： Context and attention mechanisms improve small target detection （ Attached thesis download ）
  

• Double-Head： Innovation on the detection head , Improve accuracy （ Attached is the original paper download ）

• Produced by Haikang Research Institute ： Unsupervised target detection with scene adaptive concept learning （ Attached thesis download ）
  

• New technology ： Efficient self supervised visual pre training , There is no need to worry about partial occlusion ！
  

• VS Code Supports configuring remote synchronization

• Improved shadow suppression for illumination robust face recognition
  

• Text driven for creating and editing images （ With source code ）
  

• Based on hierarchical self - supervised learning, vision Transformer Scale to gigapixel images
  

• Top frame ： Industrial testing , A semi supervised method based on difference and commonness is applied to image surface defect detection
  

• CLCNet： Rethink integrated modeling with classified confidence networks （ With source code download ）
  

• YOLOS： Rethink through target detection Transformer（ With source code ）
  

• Industrial testing ： Feature fusion based on dense scale & Pixel level unbalanced learning framework （ Paper download ）
  

• Fast YOLO： For real-time embedded target detection （ Attached thesis download ）

• Institute of computer vision ：AI Deployment and industrial landing learning （ The article is longer , Recommended collection