Pixel SkelNetOn - CVPR 2020

Organized by ilkedemir - Current server time: Jan. 21, 2021, 3:50 p.m. UTC


June 1, 2020, midnight UTC


March 18, 2020, midnight UTC


Competition Ends
June 7, 2020, midnight UTC

Pixel SkelNetOn

As the most common data format for segmentation or pixel-wise classification neural network models, our first domain poses the challenge of extracting the skeleton pixels from a given shape in an image. The participants need to overcome fundamental problems like class imbalance, global structure search, and robustness constraints while reducing the given shapes to clean skeleton pixels. Although the output will not be a true geometric representation, it is easier to convert the skeleton pixels to a vector format. We expect the challengers to provide results in terms of the accuracy better than the current best skeleton extraction from images in the system. This will be a binary classification problem to detect the skeleton pixels for a given shape image.

For details about other SkelNetOn challenges and the workshop: https://sites.google.com/view/dlgc-workshop-cvpr2020/home

Please refer to the following paper if you participate in this challenge or use the dataset for your approach:

  title={Skelneton 2019: Dataset and challenge on deep learning for geometric shape understanding},
  author={Demir, Ilke and Hahn, Camilla and Leonard, Kathryn and Morin, Geraldine and Rahbani, Dana and Panotopoulou, Athina and Fondevilla, Amelie and Balashova, Elena and Durix, Bastien and Kortylewski, Adam},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops},


Generating skeleton images from the given shape images can be posed as a pixel-wise binary classification task, or an image generation task. This makes it possible to evaluate performance by comparing a generated skeleton image, pixel by pixel, to its ground truth skeleton image. Such a comparison automatically accounts for common errors seen in skeleton extraction algorithms such as lack of connectivity, double-pixel width, and branch complexity. To minimize the effects of class imbalance, the evaluation is performed using the F1 score.

Each input is a binary image of a shape. You must predict a skeleton for the input, which is a binary image of the same height and width as the input image. 

There are 2 phases:

  • Phase 1: Development phase. We provide you with labeled training dataset and unlabeled validation dataset. You can submit your predictions on the validation data to CodaLab. You will receive feed-back on your performance on the validation set. The performance of your LAST submission will be displayed on the leaderboard.
  • Phase 2: Final phase. The unlabeled testing dataset will be released. You can submit your predictions on the testing dataset to CodaLab. Your performance on the test set will appear on the leaderboard when the organizers finish checking the submissions.

You only need to submit the prediction results (no code). However you need to submit your a short paper of 4 pages before April 10th to be eligible for the final phase. We will evaluate your methodology and your results in parallel. Paper submission is open at https://cmt3.research.microsoft.com/DLGC2020 and please use the CVPR paper template.

The submissions are evaluated using the F1 score.


This challenge is governed by SkelNetOn Rules. For academic use of the datasets within and outside this competition, please cite the following papers.
[1] I. Demir, C. Hahn, K. Leonard, G. Morin, D. Rahbani, A. Panotopoulou, A. Fondevilla, E. Balashova, B. Durix, and A. Kortylewski. "SkelNetOn 2019: Dataset and Challenge on Deep Learning for Geometric Shape Understanding." In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. 2019.
[2] A. M. Bronstein, M. M. Bronstein, A. M. Bruckstein, R. Kimmel, Analysis of two- dimensional non-rigid shapes, Intl. J. Computer Vision (IJCV), Vol. 78/1, pp. 67-88, June 2008.
[3] Sebastian TB, Klein PN, Kimia BB (2004) Recognition of shapes by editing their shock graphs. IEEE Trans Pattern Anal Mach Intell 26(5):550–571
[4] Leonard, K., Morin, G., Hahmann, S., & Carlier, A. (2016). A 2D shape structure for decomposition and part similarity. 2016 23rd International Conference on Pattern Recognition (ICPR), 3216-3221.



Start: March 18, 2020, midnight

Description: Development phase: Please submit your results on the validation set.


Start: June 1, 2020, midnight

Description: Final phase: Please submit the output of your approach on the provided test set.

Competition Ends

June 7, 2020, midnight

You must be logged in to participate in competitions.

Sign In
# Username Score
1 digitalspecialists 0.8840
2 opanichev 0.8001
3 DeepBlueAI 0.7900