How to back up?

What you request would require an (A)IP software that allows saving a 'scene'. By scene I think of a feature to save the current work's state together with the history of what was done so far. Ideally, we would then be able to re-open an (A)IP project after weeks ... or month again to try a new approach.

The tricky and possibly unsolvable requirement would be the history part: How could an (A)IP software store all steps / layers done so far, which happen to reside in RAM, cache and/or swapspace .... including the software specific features to allow navigating, undoing and redoing whatever ... after the first re-open of that project in weeks or months?

~ * ~

Example: Siril

I really love the many realtime options for image processing Siril offers: I can try out and revoke, undo and redo everything from 'open file' to the current state. But when I left Siril there is no way anymore to re-open that file again ... and go back in history for some (or more) steps.

The best (most defensive) I can do is to save the very first steps (crop + stretch + photometry) as 'file-v1.fit' for this(!) state and than continure tweak other things and save this state as 'file-v2.fit' and so on. This is similar to what you do already, I guess.

I would not (more or less) post process an image to it's end – until it meets my expectations – and then(!) save it as 'file-v1.fit' , because there is no option to go back in history. Thus I save a 'history version' at some point I am accustomed to.

I believe, there is no other way in Siril: Saving intermittend processing steps to which I can return later, when needen plus keeping the logfile so I at least can read what I did before, when I re-open such a project to rework it.

~ * ~

Example: Gimp

Gimp has layers. I developed an individual naming convention for my layers and layer groups. It's a matter of discipline to follow this individually thought out naming convention. After a long break from this one project, I re-opened it for the first time again and was very happy about my layers naming convention which gave good hints and/or precise parameters and values for what I had done at that past time.

Then I can switch off the visibiliy of one or more layers, copy them to a new layer version and start over. Because of my strictly hierarchical ordered layers, I am able to rather quickly re-catch visually what I had in mind 'then' and what I did[i].[/i]

It works for me, tough I only occasionaly use Gimp, e.g. in the context of star removal (StarNet) and than getting them back in.

On Pixinsight you can simply save the project.

That includes every image you have open with its full processing history and every process you have instantiated as an icon on a desktop.

The process history is a very powerful tool in its own right. If you have applied 10 things, you can clone the image (to save it for reference) and then revert to e.g. step 6. Or you can apply the same process history to a different original image, to compare the results.


When I know I will be working on something for weeks, I use a project.

As mentioned earlier, Pixinsight projects are very handy to keep the history of what you did.  When I'm actively working on a project, I always use projects and save them periodically.  This also lets me switch between different projects as I'm often working on more than one target at a time. 

One problem with projects, is that they're fairly inefficient with their data storage.  Even though PI uses compression, it basically saves a copy of each revision step in a given image.  If you do lots of tweaks like I do, it can really start to add up. 

The process I've ended up with is that when I'm "done" with an image, I "flatten" the individual image windows by cloning them and deleting the original image.  The cloned images keep a record of all the steps performed to create the image, but you loose the ability to roll back to an earlier state.  I make sure I keep copies of everything that went into creating the individual images like all the masks and whatnot.  This lets me at some future point to go back and recreate things if I need to.  I then save the original subs, the generated masters, and the project file onto my NAS for archiving.

If I revisit an image after some time, I'm usually adding data or taking advantage of improved pre-processing enhancements, so rolling back isn't needed, but I do want a record of the steps I took to create the original images.

As a relative newbie I still can't get the hang of using projects, so what I do is:

- open the original integrated image and immediately clone it. Close the original and rename the clone "Master"

- perform the first processing step, eg SPCC. When completed, rename the image identifier as "Master_SPCC", save the image as an xisf, produce a clone, then close the file. The clone's identifier will be "Master_SPCC_clone"

- perform the next processing step on that clone, eg DBE. When done, modify the identifier to "Master_SPCC_DBE", save as a new file, produce a new clone, then close the original image.

Continue for each step.

By doing this I have a saved version of each step, with the identifiers showing all the processing steps up to that image, eg "Master_SPCC_DBE_BlurX_GHS_NoiseX……" While this process doesn't show what parameters I used in each step, at least I can go back to any of the saved images, clone the clone, and try something new.

At the end when I have my final image I'll generally delete most of the saved copies, except for the early ones with SPCC, DBE and cropping done, as they're likely to be unchanged in any new processing.


Frank