Differences between revisions 12 and 13
Revision 12 as of 2008-07-01 01:05:00
Size: 5786
Comment:
Revision 13 as of 2008-07-23 03:17:03
Size: 5802
Comment:
Deletions are marked like this. Additions are marked like this.
Line 13: Line 13:
'''A:''' As of June 20th 2008 the answer is '''A:''' As of July 22nd 2008 the answer is
Line 20: Line 20:
e2boxer.py *.mrc --auto=db --writedb --writeimages e2boxer.py *.mrc --auto=db --write_coord_files --write_box_images

NOTE that this is a wiki and if your question is not here then simply add it. An email will be sent to all people who are subscribed to this page letting them know that the page has been changed and your question will be answered.

  • [#q1 How do I use e2boxer to automatically box many images (potentially hundreds) from the command line?]
  • [#q2 What are Dynapix and Anchor for?]
  • [#q3 Is there an iterative centration option like in boxer?]
  • [#q4 The boxes around the particles can be quite colorful white, green, yellow, black; can you give me a translation to this color coding?]
  • [#q5 Why is it creating all those *.box.bak files ? Are they important?]
  • [#q6 What is the plot in the advanced tab?]

Anchor(q1) Q: How do I use e2boxer to automatically box many images (potentially hundreds) from the command line?

A: As of July 22nd 2008 the answer is

First load up 3-10 images in the interface and use the tools to do autoboxing. Once everything looks fine and the automated boxing is behaving nicely in the interface click 'done' - this will save the autoboxing parameters to a local database.

Then quit to the command line and run something like this (make sure you specify the output you want)

e2boxer.py *.mrc --auto=db --write_coord_files --write_box_images

What's really happening is the contents of the database are being examined to autobox all of the images. When you do use the --auto=db option from the command line it uses the most recently modified/created/used autoboxer in the interface.

Some people may have both near-to and far-from focus images and may be concerned about the results of autoboxing using a single set of autoboxing parameters. At the moment there is no quick work around for this situation because you can only use one autoboxer from the database to box many images automatically from the command line - you have no choice over which 'autoboxer instance' you're using either, it is simply the most recently created/modified/used autoboxer. If you want to break your images up in close to focus and far from focus groups you'd have to put them in separate directories. But it wouldn't be surprising if you could make a single autoboxer that worked well on both close to and far from focus images- i.e. it may not be necessary to break them up at all.

Anchor(q2) Q: What are Dynapix and Anchor for ?

A: As of June 20th 2008 the answer is

  • Dynapix is a mode of operation where e2boxer does automatic picking in real time based on what 'references' the user adds (or deletes, or moves).
  • The presence of the anchoring options is debated - it means that a reference added is not used to refresh the correlation image - instead it is only used to refresh the automatic picking parameters. That may sound confusing - for the time being you could probably ignore it, or you could try and see what happens.

Anchor(q3) Q: Is there an iterative centration option like in boxer ?

A: As of June 20th 2008 the answer is

e2boxer does its own centering by default. Currently the user has no control over it. This could change if people demand it. So far most people are happy with the centering... The details are e2boxer does autoboxing in shrunken images but then automatically recenters the boxed particle in the big (original) image using translational (integer only) alignment.

Anchor(q4) Q: The boxes around the particles can be quite colorful white, green, yellow, black; can you give me a translation to this color coding ?

A: As of June 20th 2008 the answer is

  • green - an automatically selected box
  • white - a manually specified box
  • black - a manually specified box that is being used as a reference for autoboxing
  • yellow - the currently selected box, for visual reference only, there are some bugs associated with this color which need fixing

Anchor(q5) Q: Why is it creating all those *.box.bak files ? Are they important?

A: The '.bak' stands for backup - when you tell e2boxer to write a .box coordinate file it will automatically create the name of the output file based on the image name, and sometimes this file already exists. If it does exist it's copied and written as the '.box.bak' file, also a time stamp is included in the new file name. You can get around this by ticking the 'force' tick box in the 'Output' section in the main controller.

Anchor(q6) Q: What is the plot in the advanced tab?

A: Very broadly speaking, the plot displays an optimum correlation 'peak profile' (that is upside down). This peak profile is generated automatically using the references you have specified in the interface - specifically the local correlation peaks nearest to the references are located and the information in the local neighborhoods is condensed into single 'correlation peak profiles'. In the plot the horizontal axis represents radius, and the vertical axis is representative of pixel value ratio (relative to the local correlation maximum score). All of the peak profiles of each of the individual references are carefully condensed into an optimum profile, and this is what you see in the advanced tab. When you have 'selective' chosen the autoboxing algorithm uses two pieces of information to autobox - a correlation threshold, and also the part of the optimum profile that corresponds to the largest drop (greatest in the vertical direction). This latter parameter is shown by a vertical green bar in the plot in the advanced tab. If you choose 'more selective', then the autoboxer uses all of the optimum profile up to and including the radius that corresponds to the largest drop (in addition to the correlation threshold).

EMAN2/boxerFAQ (last edited 2009-03-13 02:56:14 by DavidWoolford)