The purpose of the DIC challenge is to supply the image correlation community with a set of images for software testing and verification. This is to include both commercial codes and university codes. The use of a common image data set removes the experimental errors associated with multiple hardware setups created by a typical, specimen-based, round-robin style test. The DIC code itself is then isolated and more easily evaluated independent of other experimental considerations. The DIC Challenge will be conducted under the auspices of the Society for Experimental Mechanics (SEM) under the direction of the DIC Challenge board. The purpose of the challenge is not to rank the existing codes, but to provide a framework in which all codes can be tested, validated and improved for use in experimental mechanics.
All information will be freely disseminated at the DIC Challenge website and open to all. While the images and results are open to all, there may be a requirement to limit the number of participating university codes. The open site still allows researchers to download the images and compare with the published results from the participating codes. All results will be posted and tied to the code used for the analysis. That is, each code used will be identified by name and tied to their results. Because of this, all analysis will be done by the code developers themselves. This removes any issues or concerns about misuse of the software by a third party.
Test images will be created both experimentally and synthetically. It is hoped that for all three stages of the DIC challenge we will be able to have both types of images. It will be the responsibility of the DIC board members to create and evaluate the images to best test and challenge the DIC codes. Details of the creation of the images will be recorded in a published paper so participants can understand how the images were created.
Answers will be supplied for some image sets (sample), but not for others (blind). The sample images will be supplied to allow the developers access to images similar to the blind tests but with known solutions to aid in the improvement of their code. The use of blind data sets is important to better mimic an actual experiment where the final answer is unknown; requiring the user to pick the “best” software settings without the aid of knowing the solution.
The DIC Challenge Board will create the DIC challenges, moderate and compile the tests, and disseminate and interpret the results. The board will consist of users of DIC software with enough expertise in the use of DIC codes to be able to fairly create image sets and evaluate the results. Code developers will not be involved on the board to remove any suspicion of unfairness or conflicts of interest.
Phillip Reu – Chair
Mark Iadicola (NIST) – co-chair
Elizabeth Jones (Sandia) – Results analysis
Evelyne Toussaint (University Clermont Auvergne, France) – Results analysis
Ruben Balcaen (PhD Student KU Leuven) – Stereo-DIC synthetic image generation
Hugh Bruck (University of Maryland) – Advisor at large
Helena Jin (Sandia) – Digital Volume Correlation (DVC) lead