ICDAR 2021 Competition on Components Segmentation Task of Document Photos


We are pleased to inform you that the results of the competition are available to all participants. The competition paper and the respective results can now be checked on arXiv.org through the following link: https://arxiv.org/abs/2106.08499. We had great work and great results for all teams. We emphasize that it was a very competitive and challenging competition, and we are happy with the result. Congratulations to all participants who made this competition a success!


For this competition, participants must employ image segmentation techniques and perform different levels of segmentation of real documents. The purpose of this competition is to evaluate the quality of image segmentation algorithms in case of documents acquired by mobile devices cameras, where several issues may influence the segmentation task, such as the distance of document to the camera, the perspective, the focus, reflection or shadow due to the environment luminosity, place where document is, document texture and background, among others. In addition, we are proposing in this competition different levels of segmentation tasks, yielding to the following challenges in the competition.


  • 1st Challenge: Document Boundary Segmentation
  • 2nd Challenge: Zone Text Segmentation
  • 3rd Challenge: Signature Segmentation


To evaluate the methods, the following similarity metrics will be considered:

  • Dice Similarity Coefficient (DSC);
  • Scale Invariant Feature Transform (SIFT);
  • Structural Similarity Index (SSIM).


Alejandro H. Toselli
Prof. Dr
Byron L. D. Bezerra
Prof. Dr
Celso Lopes Junior
PhD Student
Donato Impedovo
Prof. Dr
Ricardo B. Neves Junior
PhD Student