Recap: WebCT Daresbury Training

[WYVERN2742]

We had a bunch of new users using WebCT for the first time in a training session held at the STFC Daresbury campus, here are the notes and issues raised from that session:

The Good Parts

• Everyone really enjoyed the tool and ended up playing with it for quite some time, look like it's a winner!
• Apart from some specific edge cases, it worked well most of the time, including exporting / downloading and reconstruction.
• Everyone enjoyed the starvation view
• Once people got used to the UI, it seemed easy to use, not many questions apart from specific cases.
• Nobody needed help to download and run WebCT, even on Linux

The Bad Parts

• There is a regressive bug caused by fixes to geometry scaling, the act of swapping to ComputeXRayImage results in the model rotation changing the rotation axis, this results in a laminographic scan, but this doesn't translate to a change in the CIL reconstruction geometry, making results into abstract art
  • This should be reverted, and changed back to computeCTAcquisition, for now, node scaling is less important than correct rotation.
  • However, laminography support would be very very nice. This bug shows it's quite easy to implement on the simulator side, only requiring linkage to CIL reconstruction.
• There were multiple issues with input validation, X-ray tube limits failed to be indicated to users (and the user interface), resulting in invalid X-ray images, and backend errors
  • A validation framework is mostly built but should be prioritised
• Many thanks to Edo for the small UI changes recommended, such as displaying the source power and renaming simulate capture to generate X-rays, and of course bringing up the lack of dropdown arrow on projection settings again
• Another UI issue is around changing detector shape. Increasing the detector size, the image preview is stretched when the detector size is increased, but this is a side effect of the image display type. Perhaps detect when the detector is larger than the old generation, clamp to original size, and fill-in the gap with a hatched gray to indicate missing area?
• In once case, the projection preview failed to generate? WebCT logs did not indicate failure, I'm entirely sure what caused this...
• There is a case of issues surrounding NO-GPU and STFC Cloud. gVXR now has automatic backend selection with createopenglcontext, GVXR on STFC cloud fails to function (even on EGL).
  • On this same topic, we should transition from TIGRE's gpu util class to gvxr's opengl API
• Higher energy bins cause graphics VRAM usage to escalate, running into maximum texture sizes. These limits need to be exposed by gVXR for limit setting. Bin count reduction should also be performed
• The sample upload/selection are problematic. New uploaded samples should be automatically added. Existing samples should be more visible. Supports for phantoms is required due to the currently minimal state of WebCT's default STL offerings.
• Filtration needs to support custom elements and multiple filters. Also make no filter clearer.
• It would be very nice for the Material Library to better support element selection, using something like chemdoodle's periodic table canvas
• Iterative reconstruction methods really need a better UI
• Multiple times people have wanted to stop simulating or reconstruction, but that's currently impossible.

[WYVERN2742]

Understandably I mostly jotted down the issues / rough patches people had 😅