$50k First Letters Prize + Extending/Increasing the Segmentation Tooling Prize $35k ➡️ $45k
More prizes!
$50,000 First Letters Open Source Prize
We’re announcing $50,000 in new prizes for finding the first letters in the scrolls! $40,000 for the first team that uncovers 10 letters within a single area of 4cm2 from Scroll 1 or 2, and open sources their methods and results (after winning the prize). And a $10,000 prize for the second team to reach this milestone.
The purpose of this prize is to close the gap between the Ink Detection Prize on Kaggle, and the much more difficult Grand Prize. We’ll learn a lot from this: what segmentation and flattening techniques work best; what ink detection methods work well in the full scrolls, and other things we can’t predict now.
A submission consists of an image, clearly showing at least 10 legible letters in a continuous area from Scroll 1 or 2, no larger than 4cm2. Words and lines have to be clearly visible. For example, the following are areas within the fragments (all 4cm2) that have respectively ~22, ~17, and ~10 legible letters in them:
Submissions are subject to the same review process as the Grand Prize. More details can be found here.
Extending/Increasing the Segmentation Tooling Prize $35k ➡️ $45k
We announced the Segmentation Tooling Prize to encourage open source development of tools to segment the scrolls. To give teams more time to present their best work, we’re extending the deadline from May 15th to June 14th (same as the Ink Detection Prize).
We’re also increasing the prize pool by adding another $10k prize. The prizes are now:
2x $10,000
5x $5,000
Good segmentation is critical to reading the scrolls. We encourage everyone to listen closely to our contractors who are creating open source segments, and who are logging their pain points and feature requests in this doc.
Remember: tools that actually get used by our segmenters or the community will have the highest chance of winning.
Community news
The University of Kentucky researchers released a monster segment, which was created with a new tool, Quick Segment. Download instructions are here.
Ben (@Hari_Seldon on Discord) made a tutorial for how to do segmentation using Volume Cartographer.
Ben also made a very thick surface volume that spans two sheets of papyrus, with an air gap between them.
James made a tutorial for how to explore the scrolls using 3D Slicer and Slicer Morph.
Moshe Levy added support for streaming in thumbnails of slices from the data server.
Santiago Pelufo has been looking at the scrolls in Blender.
Brett Olsen has been working on detecting papyrus vs air gaps.
RICHI has been working on a new segmentation algorithm, though it hasn’t been publicly released yet!