AI program for culling?

[RichF]

Like the others, I try to reduce and minimize the number of times I go through images. My initial pass is to identify discards, identify a pool of potential selects, and move the borderline images out of the way so my second pass is only potential selects. I don't magnify images to look at details on the first pass.

On the second pass, I want to identify selects to be edited - and this includes the one from a group of similars or both vertical and horizontal selections. I only magnify the potential select group images as needed - a quick toggle is usually enough.

My third and final pass is actually editing. I might reject an image here because it's not sharp enough or some other flaw I had not noticed, but the intent is to just make a quick and efficient edit for the intended use. For events, I have a lot more images than for fine art so the time is allocated differently on a per image basis. I might have 500 Selects from an event, and just 10-15 from a day of personal fine art photography of landscapes and wildlife.

Don't underestimate the value of keywords. I keyword every photo at ingest, and then add additional keywords and captions to Selects that are going to be edited. It's a lot easier to capture keywords on the front end - even if they are somewhat generic by subject and location. Photo Mechanic does this very well.

I have with the Z9, the number of images I toss on the initial pass is much lower than with my old D850 (I keep more - many fewer images are OOF and cut off bodies is the about the same). Good news, sort of, but since the frame rate is higher (8, 10 or 20 FPS) than with the D850 (around 7 I think), I have many more keepers. The challenge becomes picking better similar images, BIF is somewhat easier due to wing and head position.

It would be nice if there was a program that would go through my images, identify images that are now sharp or where the body is cut off - place these in subdirectory "Possible deletes". Then find similar images and place them in a stack.

 

[Marsha Jane]

I have with the Z9, the number of images I toss on the initial pass is much lower than with my old D850 (I keep more - many fewer images are OOF and cut off bodies is the about the same). Good news, sort of, but since the frame rate is higher (8, 10 or 20 FPS) than with the D850 (around 7 I think), I have many more keepers. The challenge becomes picking better similar images, BIF is somewhat easier due to wing and head position.

It would be nice if there was a program that would go through my images, identify images that are now sharp or where the body is cut off - place these in subdirectory "Possible deletes". Then find similar images and place them in a stack.

Nice news about the Z9! I do wish that there were an AI program to identify the duds first off.

 

[EricBowles]

It would be nice if there was a program that would go through my images, identify images that are now sharp or where the body is cut off - place these in subdirectory "Possible deletes". Then find similar images and place them in a stack.

If they can judge sharpness, eyes, and facial expression, I suspect they will be evaluating some aspects of wildlife in the future. It's tough and can be species specific to identify the best shots, but less desirable images such as when the bird is facing away from the camera could be weeded out with technology. The Zenfolio product was interesting because you can weight the amount of emphasis to place on each criteria. At the very least, it can identify and discard all your soft images. I need to give it a test for landscapes.

 

[Marsha Jane]

If they can judge sharpness, eyes, and facial expression, I suspect they will be evaluating some aspects of wildlife in the future. It's tough and can be species specific to identify the best shots, but less desirable images such as when the bird is facing away from the camera could be weeded out with technology. The Zenfolio product was interesting because you can weight the amount of emphasis to place on each criteria. At the very least, it can identify and discard all your soft images. I need to give it a test for landscapes.

Please keep us posted!

 

[joe43]

For my Z9 images I use Nikon NX Studio to view and cull my images. I find it fast, and when needed I can use the focus square function to see exactly where the focus square was on the image when I took the picture. This sometimes helps me decide which image(s) to keep if I have similar ones captures with a burst.

 

[Marsha Jane]

For my Z9 images I use Nikon NX Studio to view and cull my images. I find it fast, and when needed I can use the focus square function to see exactly where the focus square was on the image when I took the picture. This sometimes helps me decide which image(s) to keep if I have similar ones captures with a burst.

Interesting and useful point! Thanks!

 

[ButlerKid]

A different perspective..... Perhaps consider the following:

1. YOU took all those images! Why? Wasn't there a reason why you took all those images? How would a computational AI program know what you were thinking?

2. Why are many of thse images unappealing to you now? Lack of sharpness? Poor exposure? No clear subject? Poor composition? Or ???? The images you don't want are GREAT feedback to your photography! What skills do you need to work on? Instead of avoiding those images, LEARN from them.

3. Do you have multiple images that are redundant? How can you modify you approach to photograph to get a higher percentage of "keepers"?

4. I'm sure you and others can add to this list.....but I don't believe AI is the answer................. YMMV...........

 

[RichF]

A different perspective..... Perhaps consider the following:

1. YOU took all those images! Why? Wasn't there a reason why you took all those images? How would a computational AI program know what you were thinking?

2. Why are many of thse images unappealing to you now? Lack of sharpness? Poor exposure? No clear subject? Poor composition? Or ???? The images you don't want are GREAT feedback to your photography! What skills do you need to work on? Instead of avoiding those images, LEARN from them.

3. Do you have multiple images that are redundant? How can you modify you approach to photograph to get a higher percentage of "keepers"?

4. I'm sure you and others can add to this list.....but I don't believe AI is the answer................. YMMV...........

1. Looking for the "best" one. Not asking AI to find the best one, just help me weed out the junk. only asking AI to help me find sharp ones and ones that I did not clip a wing, leg, tail, ...

2. no - unappealing only because I want the best of the best. % tossed is not unreasonble based upon what I hear others say.

3. Yes. Fine differences are important.

4. AI to help the initial cull, not final selection

 

[ButlerKid]

1. Looking for the "best" one. Not asking AI to find the best one, just help me weed out the junk. only asking AI to help me find sharp ones and ones that I did not clip a wing, leg, tail, ...

2. no - unappealing only because I want the best of the best. % tossed is not unreasonble based upon what I hear others say.

3. Yes. Fine differences are important.

4. AI to help the initial cull, not final selection

Hopefully the OP will consider them. I have learned more from my failures and out takes......For me, learning WHAT makes the "fine differences" enables me to achieve it in camera......

BTW - my keeper rate is about 1%...... I expect a LOT from myself! LOL!

 

[RichF]

Hopefully the OP will consider them. I have learned more from my failures and out takes......For me, learning WHAT makes the "fine differences" enables me to achieve it in camera......

BTW - my keeper rate is about 1%...... I expect a LOT from myself! LOL!

I try to keep around 10-20%, not that I most of these will ever see the light of day. I keep them just in case ..

 

[Marsha Jane]

1. Looking for the "best" one. Not asking AI to find the best one, just help me weed out the junk. only asking AI to help me find sharp ones and ones that I did not clip a wing, leg, tail, ...

2. no - unappealing only because I want the best of the best. % tossed is not unreasonble based upon what I hear others say.

3. Yes. Fine differences are important.

4. AI to help the initial cull, not final selection

Absolutely agree!!!

 

[Marsha Jane]

A different perspective..... Perhaps consider the following:

1. YOU took all those images! Why? Wasn't there a reason why you took all those images? How would a computational AI program know what you were thinking?

2. Why are many of thse images unappealing to you now? Lack of sharpness? Poor exposure? No clear subject? Poor composition? Or ???? The images you don't want are GREAT feedback to your photography! What skills do you need to work on? Instead of avoiding those images, LEARN from them.

3. Do you have multiple images that are redundant? How can you modify you approach to photograph to get a higher percentage of "keepers"?

4. I'm sure you and others can add to this list.....but I don't believe AI is the answer................. YMMV...........

Interesting questions to ponder, but the camera itself uses AI, so why not use AI to assist in culling?

 

[RockvilleBob]

I am in the early stage of evaluating AfterShoot. So far I am finding it a big help since I shoot many BIF shots and between light, focus, composition, etc. only a few are really good. Wildlife photography in ways is the opposite of event photography, like a wedding. Event photography is staged so that photographers are dealing with multiple images of a staged scene, The wedding culling job, for example, might be going through a wedding group picture and tossing those with eyes closed, heads turned, facial expressions, smiles etc. There may be many images of the same scene. Wildlife is not staged and therefore results in more images with often fewer "duplicate" shots.
After doing this I took a past outing that I culled manually using FastRawViewer and Bridge. Out of about about 2600 shots I eventually wound up with 5 that I rated good enough to keep and different from shots I already have from previous outings. This took a fair amount of time.
With Aftershoot I wound up with 196 shots to keep with less than 5 minutes of my time. Going through the 196 to compare with the 5 I marked manually I found all five were in the 196. The program itself tosses nothing and yielded about 800 scenes some with identical images, classified as out of focus or fuzzy. It was a simple job, less than 5 minutes, to go from 800 to 196. If I can get from 2600 to 196 with 5 minutes of work - that is a winner for me.
I did not go through the 196 to kick files I already hd and would not save. This would not take much time.
Bottom Line: With a modification to reverse selection from a staged event to wildlife photography I found AfterShoot to be a very big help for my photography.
THIS IS BASED ON ONE SET OF 2600 IMAGES FROM ONE OUTING. I WILL BE DOING MORE TESTING BEFORE MY 30 DAY TRIAL EXPIRES.
IT DOES TAKE SOME COMMITTED TIME TO GET TO KNOW THE PROGRAM AND IT PROBABLY HELPS IF YOU ARE A LITTLE ON THE TECHY SIDE.

 

[Marsha Jane]

I am in the early stage of evaluating AfterShoot. So far I am finding it a big help since I shoot many BIF shots and between light, focus, composition, etc. only a few are really good. Wildlife photography in ways is the opposite of event photography, like a wedding. Event photography is staged so that photographers are dealing with multiple images of a staged scene, The wedding culling job, for example, might be going through a wedding group picture and tossing those with eyes closed, heads turned, facial expressions, smiles etc. There may be many images of the same scene. Wildlife is not staged and therefore results in more images with often fewer "duplicate" shots.
After doing this I took a past outing that I culled manually using FastRawViewer and Bridge. Out of about about 2600 shots I eventually wound up with 5 that I rated good enough to keep and different from shots I already have from previous outings. This took a fair amount of time.
With Aftershoot I wound up with 196 shots to keep with less than 5 minutes of my time. Going through the 196 to compare with the 5 I marked manually I found all five were in the 196. The program itself tosses nothing and yielded about 800 scenes some with identical images, classified as out of focus or fuzzy. It was a simple job, less than 5 minutes, to go from 800 to 196. If I can get from 2600 to 196 with 5 minutes of work - that is a winner for me.
I did not go through the 196 to kick files I already hd and would not save. This would not take much time.
Bottom Line: With a modification to reverse selection from a staged event to wildlife photography I found AfterShoot to be a very big help for my photography.
THIS IS BASED ON ONE SET OF 2600 IMAGES FROM ONE OUTING. I WILL BE DOING MORE TESTING BEFORE MY 30 DAY TRIAL EXPIRES.
IT DOES TAKE SOME COMMITTED TIME TO GET TO KNOW THE PROGRAM AND IT PROBABLY HELPS IF YOU ARE A LITTLE ON THE TECHY SIDE.

Amazing! Thanks for your work and report!

 

[RichF]

I am in the early stage of evaluating AfterShoot. So far I am finding it a big help since I shoot many BIF shots and between light, focus, composition, etc. only a few are really good. Wildlife photography in ways is the opposite of event photography, like a wedding. Event photography is staged so that photographers are dealing with multiple images of a staged scene, The wedding culling job, for example, might be going through a wedding group picture and tossing those with eyes closed, heads turned, facial expressions, smiles etc. There may be many images of the same scene. Wildlife is not staged and therefore results in more images with often fewer "duplicate" shots.
After doing this I took a past outing that I culled manually using FastRawViewer and Bridge. Out of about about 2600 shots I eventually wound up with 5 that I rated good enough to keep and different from shots I already have from previous outings. This took a fair amount of time.
With Aftershoot I wound up with 196 shots to keep with less than 5 minutes of my time. Going through the 196 to compare with the 5 I marked manually I found all five were in the 196. The program itself tosses nothing and yielded about 800 scenes some with identical images, classified as out of focus or fuzzy. It was a simple job, less than 5 minutes, to go from 800 to 196. If I can get from 2600 to 196 with 5 minutes of work - that is a winner for me.
I did not go through the 196 to kick files I already hd and would not save. This would not take much time.
Bottom Line: With a modification to reverse selection from a staged event to wildlife photography I found AfterShoot to be a very big help for my photography.
THIS IS BASED ON ONE SET OF 2600 IMAGES FROM ONE OUTING. I WILL BE DOING MORE TESTING BEFORE MY 30 DAY TRIAL EXPIRES.
IT DOES TAKE SOME COMMITTED TIME TO GET TO KNOW THE PROGRAM AND IT PROBABLY HELPS IF YOU ARE A LITTLE ON THE TECHY SIDE.

if this is repeatable, this is great news

 

[RichF]

I am testing Aftershoot. Three folders

All images are from a day of heavy shooting at Bosque

1. Images (yet) unsorted
2. Images I decided in pass 1 are keepers (sharp, birds do not overlap, wings/heads/body parts are not cut off)
3. Images to delete (OOF, birds overlap, in a few case too many duplicates)

Will post results

 

[Marsha Jane]

I am testing Aftershoot. Three folders

All images are from a day of heavy shooting at Bosque

1. Images (yet) unsorted
2. Images I decided in pass 1 are keepers (sharp, birds do not overlap, wings/heads/body parts are not cut off)
3. Images to delete (OOF, birds overlap, in a few case too many duplicates)

Will post results

Great — I am presently mulling over my images from Bosque, too!

 

[clementstevens]

Having just returned from Bosque del Apache with thousands of photos, I am getting bogged down culling. Is there an AI program out there to assist in culling wildlife (specifically, bird) images? I am looking for one that I can use either on my iPad iPro, or Mac desktop. Thank you!

No, this requires human intervention

 

[RichF]

Aftershoot summary. Keep in mind that this is from one day at Bosque. I was crane pond just before you reach the visitors center and then the duck pond on the UNM campus in Socorro. I had previously sorted some of the images of the day

I (luckily) had saved my "to delete" images and my keepers (plus there are few I had yet to sort).

The "to delete" included OOF image, blurs which in my opinion did not work, images where the bird (generally wing) was cut off, images where the pose did not work (i.e., silhouettes) images were the image was just too crowded with birds in front of each another, etc.

Total images 1,592 from my To Delete Folder
- Selected (5 stars) - 522 images. A lot of these were flagged for deletion because the wings, legs were clipped off. in other cases birds were superimposed. Or I simply did not like the wing position.
- Highlights (4 stars) - 162 images. similar to above but I found images that tossed due to being slightly OOF, were kept.
- Blurred (2 stars) - 731 images.


Total Images - 1,165 images from my Keeper (pass 1) folder
-Selected (5 stars) - 469 images. I would have expected more images in this group
- Highlights (4 starts) - 147 images. SImilar to selected but not as tact sharp.
- Blurred (2 stars) - 469 images. Includes motion blur. I don't expect an program to id keepers in this category. Flagged some images where the front bird is sharp and the one behind is not. THis is a failure IMO.
Found 211 duplicates. With 10-20 FPS I suspect that there will be duplicates.

I also ran it on a small number of images I have yet to sort from this day. I have yet to analyze these assignments.

Bottom line for me. Not sure if Aftershoot is worth the effort or cost.

 

[MikeInMI]

I also just got back from Bosque del Apache with 17k images to cull. Last year I started using FastRawViewer but more recently moved to Narrative Select AI. Narrative Select AI has AI features for people; i.e., it quickly lets you know if someone's eyes are closed and whether or not faces are in focus -- great for group shots of people, but of no help for wildlife. I've communicated with the company and they have said that they eventually hope to be able to have features for animals.

Nonetheless, I still find it the best tool I've seen to get through images quickly. I give the close ones a one star rating in Narrative Select, and then use its "ship to" feature to load just the selects directly into Lightroom. Best of all, the product is free to use for up to four projects/month with all AI features, but you can use the non-AI features free as often as you'd like. I've gotten better at being pretty ruthless in my culling (I had to). On my first pass of my Bosque pictures I went from 17k to 160, and then used Lightroom to get down to 53 to show my photography buddies and 24 to show the rest of the world that does not have the attention span and interest to marvel at my great work! LOL. Good luck!

 

[RichF]

I also just got back from Bosque del Apache with 17k images to cull. Last year I started using FastRawViewer but more recently moved to Narrative Select AI. Narrative Select AI has AI features for people; i.e., it quickly lets you know if someone's eyes are closed and whether or not faces are in focus -- great for group shots of people, but of no help for wildlife. I've communicated with the company and they have said that they eventually hope to be able to have features for animals.

Nonetheless, I still find it the best tool I've seen to get through images quickly. I give the close ones a one star rating in Narrative Select, and then use its "ship to" feature to load just the selects directly into Lightroom. Best of all, the product is free to use for up to four projects/month with all AI features, but you can use the non-AI features free as often as you'd like. I've gotten better at being pretty ruthless in my culling (I had to). On my first pass of my Bosque pictures I went from 17k to 160, and then used Lightroom to get down to 53 to show my photography buddies and 24 to show the rest of the world that does not have the attention span and interest to marvel at my great work! LOL. Good luck!

I need a lesson from you on culling. I seldom get below 10%

 

[Marsha Jane]

Aftershoot summary. Keep in mind that this is from one day at Bosque. I was crane pond just before you reach the visitors center and then the duck pond on the UNM campus in Socorro. I had previously sorted some of the images of the day

I (luckily) had saved my "to delete" images and my keepers (plus there are few I had yet to sort).

The "to delete" included OOF image, blurs which in my opinion did not work, images where the bird (generally wing) was cut off, images where the pose did not work (i.e., silhouettes) images were the image was just too crowded with birds in front of each another, etc.

Total images 1,592 from my To Delete Folder
- Selected (5 stars) - 522 images. A lot of these were flagged for deletion because the wings, legs were clipped off. in other cases birds were superimposed. Or I simply did not like the wing position.
- Highlights (4 stars) - 162 images. similar to above but I found images that tossed due to being slightly OOF, were kept.
- Blurred (2 stars) - 731 images.


Total Images - 1,165 images from my Keeper (pass 1) folder
-Selected (5 stars) - 469 images. I would have expected more images in this group
- Highlights (4 starts) - 147 images. SImilar to selected but not as tact sharp.
- Blurred (2 stars) - 469 images. Includes motion blur. I don't expect an program to id keepers in this category. Flagged some images where the front bird is sharp and the one behind is not. THis is a failure IMO.
Found 211 duplicates. With 10-20 FPS I suspect that there will be duplicates.

I also ran it on a small number of images I have yet to sort from this day. I have yet to analyze these assignments.

Bottom line for me. Not sure if Aftershoot is worth the effort or cost.

I tried AfterShoot last night. I have nothing to compare it to, as I have not used any culling program before, but here were my results. I had 2081 images from 6 days of shooting in Bosque. I used the default settings of AfterShoot. It selected 960 5 star, 849 duplicates, 12 warnings, 168 Highlights and 104 Blurred (there were many more). It took about 25 minutes for these results. I found it overly "generous" with the 5 star ratings. It identified as sharp a lot of images in which the scenery was sharp, but BIF (in front of scenery) were OOF. It gave a few 2-star ratings to images that I liked - these were small birds (ducks, sandpipers) in highly reflective water. It was good at eliminating a lot of duds. I liked the layout of the images for culling and the preliminary ratings, and I was able to go through everything in about 1.5 hours, and have the keepers (about 480) saved automatically for further review later in a separate file on my hard drive, without eliminating the photos in Apple Photo on Mac. I plan to try Fast Raw viewer to go through these. AfterShoot is relatively costly for Canadian hobbyists like myself ($120 USD annually after the free trial), and it's clear that it's not designed for wildlife, but it did help to get me going on the task. I am no techie, and I found it to be user-friendly. Support/reply was prompt for questions.

 

[Marsha Jane]

I need a lesson from you on culling. I seldom get below 10%

I am going to give Narrative Select a try, too, on the "keepers" (almost 500) from AfterShoot.

 

[RockvilleBob]

This is in reference to AfterShoot:

• I feel it is a personal judgement. How well AfterShoot works depends a lot on the outing or images in a file. When I shot BIF I get a lot of similar images from shooting bursts at 20 fps. This is also the case with if I stay on a bird like a kingfisher and take lots of shots diving from a perch. AfterShoot, for the cases I've tried, puts similar shots in a group and uses it's algorithm to select the best.
• However if my photo outing produces all or many distinct images then you get a lot of distinct images out of the program and AfterSHoot would be of less utility.
• AfterShoot also is designed to eliminate selected images by hitting a programmed key, by default "X" - the way a wedding photographer would work. Unfortunately this is backwards for wildlife shooting. The problem is to go through the selected images and easily flag those you want to keep - not delete - since I delete a lot more than I keep - the opposite of event photography. Using AfterShoots colors and stars does let you do this without modification. My work-around was to go into settings and make the maximum rating the AfterShoot culling process yields 4 stars. I find it relatively simple to go through the resulting 4 star array of AfterShoot selected images and hit 5 for the ones I may want to look at further.
• When the images are saved I need to select them by rating - 5 stars - on the save page

Using this modified method I have found AfterShoot a big help for the case where I have many images that look similar, that is the case for me when I go out to shoot BIF. For example coming back from shooting flying eagles at Conowingo and having 1000 or so eagle shots on a good day. If ( I wish) I was on a safari and came back with a 1000 shots of different mammals in different positions the AfterShoot would be less useful.

I also use AfterShoot to ingest images from my card. When I finish the AfterShoot culling and my 5 star cull I just save the resulting 5 star images on my computer.

I am still evaluating AfterShoot but so far, using this modified approach of setting 4 stars as the maximum for AfterShoot, going through the array and hitting 5 to only pick the one I want to consider further I find AfterShoot saves me a lot of time. Indeed I may miss an image from doing it all manually but I can get through a stack of images from a day shooting BIF in minutes an only copy the images I want to consider further on my computer.

I tend to come home with many similar images since I try to focus on BIF. For me I am thinking $10 a month is not a lot of money if it can save me hours going through lots of BIF images.

I still have 25 days to go and I plan on going out often to evaluate AfterShoot and play with my new lens.

As I stated at the outset - a personal decision. It also depends on the type of shooting on a specific set of images.

 

[Viathelens]

I am going to give Narrative Select a try, too, on the "keepers" (almost 500) from AfterShoot.

Interesting. I'm not sure, based on what you said, if it would be worth the cost. I shot almost 10,000 images in Bosque. I used the LrC travel catalog I always set up on an external hard drive when on a trip to delete the obvious bad images. It did not take too long, but probably a couple of hours over time. I do get tired of sitting at some point. I also selected some images at the same time for later consideration, I use a green border to do this. When I had culled a couple of thousand I imported the remaining images into LrC to continue culling, I like to review my images myself because sometimes I know that when I took the shot I had something specific that I planned to do with it in post and an automatic culling program would not work for that. I also know by quickly looking if I can fix something that an automatic culling program would not know about, too. I used export as a catalog to import the image metadata, the green boxes or any 5-starred images, into my desktop LrC application. The metadata simply layers over the already imported image. I'm down to 7,955 images now (Ha! Ha!). I have 60 images finished and over 1,000 marked with a green border so a ways to go yet. I'll probably end up with maybe a 100 or 150 images total that are edited. I may delete more or I may just leave them if not altogether bad shots. I like working in LrC because of the flexibility of the program and the numerous ways I can do things.