sikuli-driver team mailing list archive

[RaiMan <rmhdevelop@xxxxxx>]

** Summary changed:

- Region.highlight() fade out might influence following find ops
+ Region.highlight() fade out might influence directly following find ops

** Description changed:

+ ---- summary and workaround
+ in the given case this was the test code in principal:
+ 
+ exists(image)
+ getLastMatch().highlight(2)
+ find(image)
+ 
+ The find() did not match the same or even failed compared to the
+ previous exists().
+ 
+ The reason is the asynchronous fade out of the highlight frame, that is
+ still present on the search screenshot of the following find.
+ 
+ -- workaround:
+ exists(image)
+ getLastMatch().highlight(2); wait(1)
+ find(image)
+ 
+ the additional wait() pauses the script until the highlight frame has
+ vanished.
+ 
+ This is accepted as a bug.
+ 
+ -------------------------------------------------------------------------
+ 
  I have the following situation: I'm trying to see if one variant of an
  image exists, and if it doesn't, then I continue to the next variant.
  
  I'm using the IDE to determine the match parameters; the threshold for
  my value to appear in the matches is 0.58; below that, it will give me
  false positives. As the color does not need to match perfectly, I'm
  currently matching at 0.75. However, when I use the exists() method, it
  will return me a _false positive_, but the match is high; 0.77. This is
  incorrect and inconsistent with the find() method, which would have
  ignored this piece.
  
  The image i'm trying to match is an orange button, with white text; on
  the page, it will match a piece of a band of solid gray, the same size,
  but it has none of the details that the source image exhibits.
  
  Is this a known issue? And if so, apart from different pattern match
  criteria, can I do to avoid having this discrepancy?
  
  BTW: still on W7/64 with r930. Vision algorithm is at its default
  settings, haven't changed anything there. Can send the images by email
  on request, but I'd rather not share on a public place.

-- 
You received this bug notification because you are a member of Sikuli
Drivers, which is subscribed to Sikuli.
https://bugs.launchpad.net/bugs/1087315

Title:
  Region.highlight() fade out might influence directly following find
  ops

Status in Sikuli:
  In Progress

Bug description:
  ---- summary and workaround
  in the given case this was the test code in principal:

  exists(image)
  getLastMatch().highlight(2)
  find(image)

  The find() did not match the same or even failed compared to the
  previous exists().

  The reason is the asynchronous fade out of the highlight frame, that
  is still present on the search screenshot of the following find.

  -- workaround:
  exists(image)
  getLastMatch().highlight(2); wait(1)
  find(image)

  the additional wait() pauses the script until the highlight frame has
  vanished.

  This is accepted as a bug.

  -------------------------------------------------------------------------

  I have the following situation: I'm trying to see if one variant of an
  image exists, and if it doesn't, then I continue to the next variant.

  I'm using the IDE to determine the match parameters; the threshold for
  my value to appear in the matches is 0.58; below that, it will give me
  false positives. As the color does not need to match perfectly, I'm
  currently matching at 0.75. However, when I use the exists() method,
  it will return me a _false positive_, but the match is high; 0.77.
  This is incorrect and inconsistent with the find() method, which would
  have ignored this piece.

  The image i'm trying to match is an orange button, with white text; on
  the page, it will match a piece of a band of solid gray, the same
  size, but it has none of the details that the source image exhibits.

  Is this a known issue? And if so, apart from different pattern match
  criteria, can I do to avoid having this discrepancy?

  BTW: still on W7/64 with r930. Vision algorithm is at its default
  settings, haven't changed anything there. Can send the images by email
  on request, but I'd rather not share on a public place.

To manage notifications about this bug go to:
https://bugs.launchpad.net/sikuli/+bug/1087315/+subscriptions