Sorting out blurry. not perfectly crisp/sharp photos automatically ?

[kirk]

Is there something in iMatch or available tagging AI  to sort photos by pixel crispness vs blurriness ?

I understand those ones having bokeh  could be marked as blurry   but since I usually avoid bokeh as much as I can ( landscape scenes)  I would perfectly ok with that.

Although for  bokeh and corners blur I imagine it could be some 10% of pixels crisp,  20% crisp   tag or something


Wonder If any software  can do it currently?

[kirk]

Found this https://github.com/WillBrennan/BlurDetection2

Wonder if I can use it together with iMatch somehow?

[Mario]

This is a Python script. You can install Python and run the script to identify "blurry" images (expect many false positives) and then further process the files?
If the Python script takes names on the command line, you can run it from IMatch, handing over the names of selected files.
If the Python script process an input file with file names, you can most likely produce it with the Text Exporter in IMatch and then let the script run.

I see that the code is based on the work of another programmer, which used OpenCV and Python to implement this feature.
IMatch uses OpenCV for several purposes. Implementing blur detection would be possible but would take time - aka cost money. And the results would most likely be debatable, unless you somehow can configure this to only flag really out-of-focus images. This is not an AI and differentiating accidental blur from motion blur or bokeh is complex and probably never accurate.

Usually you weed blurry images out when doing the initial culling in IMatch or your RAW processor.
What is your workflow and why do you have so many blurry images that you need an AI or dedicated feature to identify them?
Just curious.

Feel free to add a feature request in the feature request board: https://www.photools.com/community/index.php?board=12.0
This allows other users to comment and to +1 your request to show support.

Blur detection is a research branch in the industry, and many companies throw money at this. Probably you can use a cloud service some day to do this, similar to what the IMatch AutoTagger is doing when automatically adding keywords to files.

[kirk]

I work with other people photos done in a kind of reference trips  with often bad lighting condition, done from a car  with lots of motion  blur on the very foreground  and from drones.
   Thousands of photos usually after such a trip . Photos in photogrammetry series taken to reproduce 3d scene.   Checking all those photos manually  is kind of tiresome and  having a photogrammetry series with a few  unnoticed  blurry pictures sneaked in  usually creates issue for Reality Capture soft.

[mastodon]

This would be a nice add on!

[DigPeter]

I would think it is better to do this as part of the initial weeding process, using whatever suitable app is available.

[Jingo]

I would think it is better to do this as part of the initial weeding process, using whatever suitable app is available.

Though everyone's workflow is different - I too do initial culling upon Import. 

For Camera images: in Photo Mechanic which I use to offload images from the SD cards.... it is a fast process - culling images that are just plain bad is the first step in my workflow.  From there, images get renamed/downloaded to folders and a second / ore detailed culling occurs in my RAW Editor... only final images make it to my IMatch DB.

For Phone images: in FastStone Image Viewer after copy to disk.  Quick review, delete and cull and then import to IMatch.

For any images that get into Imatch that are not great - a quick Ctrl-Delete and they are gone.