Welcome to the TIGER challenge! ¶
By: f.ciompi on Jan. 12, 2022, 9:36 a.m.
You can use this forum section to ask questions and discuss problems relateds to the TIGER challenge and to the grand-challenge.org platform.
By: misagh.naderi on Jan. 14, 2022, 8:04 p.m.
In the Rules section under the sub-heading "Data", are there supposed to be links included? There are some underlined text that seems to be links but aren't active. It would be much appreciated if you could provide some explanation/details re training datasets (i.e., size, image labels/annotations, etc.) and/or provide some relevant external links.
Also with regards to the training datasets provided for this challenge, have you added any additional annotations to the TCGA-BRCA, RUMC, and JB slides?
And are you planning to release the test sets to the public anytime after the challenge?
By: f.ciompi on Jan. 15, 2022, 7:32 p.m.
Details about the data and a lnk to download the training set can be found in the Data section.
About annotations, all slides in the WSIROIS and WSIBULK subsets contain manual annotations. WSIROIS contain annotations of multiple tissue compartments, and of lymphocytes + plasma cells; these are release both at whole-slide image level and at ROI level, where we release crops of the WSIs in PNG format and corresponding annotations, plus cell annotations in COCO format; we think that the ROI format can be useful for participants that do not have experience with handling whole-slide images but that want to plug in computer vision pipelines of detection and segmentation. WSIROIS contain slides from TCGA, RUMC and JB. In WSIBULK, coarse annotations of the regions containing tumor cells are provided along with slides. This means that non-annotated regions do not contain tumor cells, but can contain any other type of tissue. WSIBULK only contains slides from RUMC and JB.
About releasing the test set, there are no plans at the moment to do that, this is a challenge where all test data is not directly accessible by the participants, who have to upload an algorithm in the form of a docker container, and then the algorithm is run on the test data (the experimental test sets during the challenge, the final test sets at the end of the challenge).
We are preparing a number of videos to present all these things in detail, we hope we can release them next week.
By: mart.vanrijthoven on Feb. 10, 2022, 6:02 p.m.
Dear Avipul313
The entire training data comprises 106GB.
Best whishes, Mart