Creating this record of issue for posterity, though this is something we have discussed for years and something that was only possible once webGL came online. Jess has been working on this intermittently for the Spring 2022 semester, and made serious inroads over spring break.
Implementing the published approaches to contrast detection was fairly swift, though there are some bigger issues that come up with multi-resolution image tile pyramids and higher dynamic range. Some of the basic challenges we've been working through involve accuracy/precision error when points are detected mid way up the image tile pyramid, filtering/kernel algorithms that are generalizable across gymnosperm anatomies, and efficiency of algorithm operation within the web browser, etc. Jess has come up with some very nice multi-stage approaches to the resolution issue that seem to improve outcomes, and is working to implement some basic ideas we have about UI functionality, including icons to distinguish points placed automatically vs. those placed manually (or those auto points that were moved manually), and user-control over the webGL filtering.
Jess's progress, particularly over the last two weeks, has been staggering. The first live iteration of this will come with myriad caveats (works well with some anatomies - good luck!) but is very promising, not just for annual boundary placement but also for the sub-annual boundary placement.
Creating this record of issue for posterity, though this is something we have discussed for years and something that was only possible once webGL came online. Jess has been working on this intermittently for the Spring 2022 semester, and made serious inroads over spring break.
Implementing the published approaches to contrast detection was fairly swift, though there are some bigger issues that come up with multi-resolution image tile pyramids and higher dynamic range. Some of the basic challenges we've been working through involve accuracy/precision error when points are detected mid way up the image tile pyramid, filtering/kernel algorithms that are generalizable across gymnosperm anatomies, and efficiency of algorithm operation within the web browser, etc. Jess has come up with some very nice multi-stage approaches to the resolution issue that seem to improve outcomes, and is working to implement some basic ideas we have about UI functionality, including icons to distinguish points placed automatically vs. those placed manually (or those auto points that were moved manually), and user-control over the webGL filtering.
Jess's progress, particularly over the last two weeks, has been staggering. The first live iteration of this will come with myriad caveats (works well with some anatomies - good luck!) but is very promising, not just for annual boundary placement but also for the sub-annual boundary placement.
Way to go Jess, so glad to have you on the team!