Benchmarking Data Neuroglancer Skeletons are Wrong

Open tathey1 opened this issue 4 years ago • 2 comments

Downloading benchmarking notebook shows improper traces for validation_21 with trace number 2.

We printed out "swc_string" from here, and found improper looking skeleton coordinates.

We concluded that the skeleton objects on s3 for the benchmarking data are wrong. Let's replace them with good ones.

Mar 23 '21 18:03 tathey1

We narrowed down the issue to Cloudvolume's Skeleton.to_swc() and Skeleton.from_swc() functions, theorizing that calling these functions during the swc upload and download process changes the underlying topology of the swc file. See issue posted to CloudVolume here: https://github.com/seung-lab/cloud-volume/issues/466

Apr 29 '21 03:04 shreyasingh1

Stale issue message

Feb 09 '23 19:02 github-actions[bot]