disable page corner detection or easy button to reset it

[gradha]

Trying the software with a few pictures of scanned documents I got some really weird looking thumbnails. Turns out the page detection was failing hard and in some cases, like the picture below, getting really weird shapes which are a pain to manually go one through one and manually restore:

For the case of an individual picture I would like to have an additional button which "resets" the corners to their full size.

Also, since touch is a bit hard to control with fat sausage fingers, or pressing this hypothethcail reset button by mistake might destroy previous careful work, it would also be nice to have an undo/redo toolbar buttons which allowed to go through the history of the modifications (just a list of all the different corner states would be enough).

Then, on the main capture screen, where the crop button is, it would be nice to have a long press option which either showed an additional menu or directly allowed to reset the corners of all pages without having to go one by one.

The long press option might be a good idea to propagate to the rotate buttons as well, so you could batch rotate everything by 90 degrees.

[pynicolas]

Thank you for your feedback.

When opening an issue, it is generally more helpful to describe the problem you face than to propose a specific solution. Different solutions may exist for the same problem, and understanding the underlying problem is usually the most valuable information.

From your description, the underlying issue seems to be that document detection sometimes produces abnormal quadrilaterals. Screenshots don't help to understand why detection fails. The original images would be much more helpful (you can send them privately at contact@fairscan.org)

I may be wrong, but my understanding is that these images were imported rather than captured directly with FairScan. In the normal capture workflow, the live camera preview provides immediate feedback about the detected document and helps users frame it correctly before taking the picture. With imported images, that feedback step is absent.

If my understanding of the problem is confirmed, a possible direction would be to make FairScan more conservative and reject highly suspicious detections, falling back to the full image instead (similar to what already happens when no document is detected at all).