• Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
    Dedicated community for Japanese speakers
  • 한국 커뮤니티
    Dedicated community for Korean speakers
Exit
9

Face detection also detects non-human faces

Participant ,
Jan 03, 2024 Jan 03, 2024

Copy link to clipboard

Copied

It can't even distinguish between a human and a dog!? 🤣 🤦

 

LrC Face Detection #EpicFail.png

TOPICS
macOS

Views

254

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Participant ,
Jan 03, 2024 Jan 03, 2024

Copy link to clipboard

Copied

Not sure why this isn't a bug, but whatever.

 

It raises the question of why LrC's face detection hasn't been removed. Its slow, it's error rate is extremely high, and detected people/added keywords don't sync to Lightroom Cloud. When I did a search for posts on LrC face detection, the common thread was it should removed! 

 

So why is this feature still available? IMO, Adobe should pull the feature entirely or update to the detection tech Lightroom Cloud uses; it seems to be much more accurate.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jan 03, 2024 Jan 03, 2024

Copy link to clipboard

Copied

Why should it be removed? Nobody forces you to use it...

 

-- Johan W. Elzenga

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Participant ,
Jan 03, 2024 Jan 03, 2024

Copy link to clipboard

Copied

Three reasons:

  1. It's wrong so often that it just wastes people's time.
  2. All the posts from people asking how to stop it when they've accidentally triggered it. 
  3. It's a stale feature in a dying product as Lightroom Cloud is clearly Adobe's future in this area.

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Community Expert ,
Jan 03, 2024 Jan 03, 2024

Copy link to clipboard

Copied

quote

Three reasons:

  1. It's wrong so often that it just wastes people's time.
  2. All the posts from people asking how to stop it when they've accidentally triggered it. 
  3. It's a stale feature in a dying product as Lightroom Cloud is clearly Adobe's future in this area.

By @Uhl

 

I don't use face recognition myself, but I know there are plenty of people that do and are satisfied with how it performs. That is why removing it would be a mistake. If it wastes people time, then they should simply stop using it, not demand that other people should be deprived of it. If you don't like beer, should beer be banned for everybody else?

 

Your only valid remark IMHO is that it should indeed be easy to stop it if you accidentally triggered it, but that's another question than removing it altogether.

 

-- Johan W. Elzenga

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Participant ,
Jan 03, 2024 Jan 03, 2024

Copy link to clipboard

Copied

Not a "demand," just a recommendation to remove or improve the feature. That's why I prefaced it with "IMO" (in my opinion).

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Participant ,
Jan 04, 2024 Jan 04, 2024

Copy link to clipboard

Copied

Lightroom Classic face detection will suggest non-human animals (e.g., dogs) and physical objects (e.g., speaker, cycling jersey) as named people.

 

This is important to me as I have 30K photos, many of which have faces that I need to identify. I don't want to waste time reviewing objects that are clearly not people. I have also used Google Photos, and it's never suggested a non-human image as a person, so I know it's possible with today's technology. I'm hoping LrC can catch up in this area. And yes, I'm fine if I have to upload my images to the cloud to get better face detection.

 

Lightroom Classic version: 13.1 Release Camera Raw 16.1 Build [202312111226-41a494e8]
OS version: macOS 14.2.1 (23C71)

 

Reproduction Steps

Note: This is a difficult repro as it requires a large number of photos with faces and other objects.

With a large photo library/catalog opened…

  1. Select ViewPeople (shortcut letter O), or click the People button at the bottom of the gallery
  2. Name some of the detected faces
  3. Wait for Lighroom Classic to detect more faces
  4. Review the suggested matches in the Unnamed People section  


Expected result: Suggestions should always be of people, ideally ones who look similar to the suggested person.
Actual result: Some suggestions are of non-human animals and inanimate objects.

 

Examples

LrC Face Detection Object 1.png 

LrC Face Detection Object 2.png

LrC Face Detection Object 3.png

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Participant ,
Jan 04, 2024 Jan 04, 2024

Copy link to clipboard

Copied

@Rikk Flohr: Photography I'm curious why this was moved from Bugs? I thought I followed the bug report template to a T, and it's clearly not working as intended/designed.

 

What do I need to do to improve my future bug reports?

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Adobe Employee ,
Jan 04, 2024 Jan 04, 2024

Copy link to clipboard

Copied

A 'perfect' bug report is merely a report. It is not a guarantee of an acknowledgment of the existence of a bug.  Most 'perfect' reports filed by well-meaning customers are not actually bugs (about 7 out of 8).  Some of this is due to customers not understanding the definition of a bug or if a bug is involved,  whose bug it is (Adobe's, OS, third party, etc.). Others don't understand how things should work and become flummoxed when they don't know or understand what is happening. 

As explained yesterday, this is more likely a local model training issue than a bonafide bug. 

To be a bonafide bug,  the feature must behave other than as designed, have repeatable steps to failure, and have quality corroboration of the failure by others. I do not feel that the bar has been met here. 

I think it would be reasonable to request enhancement (Idea) with a title like "Improve Face Detection Model" and then include ORIGINALS of example failures. 

Caution:

In the realm of Ideas, there is also a mistaken notion that a nicely written request is a contract and that anything asked for will eventually be fulfilled. This is not the case. There is also a guideline for making a request of this nature: https://community.adobe.com/t5/lightroom-classic-ideas/how-do-i-write-a-feature-request/idi-p/123863... 

Rikk Flohr - Customer Advocacy: Adobe Photography Products

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Participant ,
Jan 04, 2024 Jan 04, 2024

Copy link to clipboard

Copied

LATEST

Fair enough. Maybe it isn't a "bug", based on that definition, but it'd be nice to get acknowledgment that there's room for improvement.

 

Apple Photos does their face detection on device and it's never given me a false positive as bad as these examples. Perhaps LrC's local model can be improved or "pre-seeded" with generic faces? I'm not an AI developer, so I'm just spit-balling, but based on my experience with other products (including Lightroom Cloud), it can be better. 

 

I've been using Lightroom Classic for over 15 years and I truly hope it'll continue to be developed and improved. If it weren't for LrC's map and GPS route tagging, I'd probably just switch to Lightroom. But I rely on the map features for photos taken with my non-GPS camera.

 

Anyway, just my feedback as a longtime customer. Hope the Lightroom Classic team can consider this as something to work on. 🙏

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Adobe Employee ,
Jan 03, 2024 Jan 03, 2024

Copy link to clipboard

Copied

False Positives and erroneous detections are part of any AI model training.  As you train your Face Detection, results will improve, but you won't eliminate them altogether. 

The detection method used by Lightroom Eco is cloud-based and requires your images all be loaded to the cloud where the detection takes place. Do you want to commit to that?

I use Face Detection in LrC on my 410K image catalog and have very few issues. The results are accurate and when I invest the time to make corrections, the results improve. 

Rikk Flohr - Customer Advocacy: Adobe Photography Products

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
Participant ,
Jan 03, 2024 Jan 03, 2024

Copy link to clipboard

Copied

quoteThe detection method used by Lightroom Eco is cloud-based and requires your images all be loaded to the cloud where the detection takes place. Do you want to commit to that?

By @Rikk Flohr: Photography

 

Yes, I just did! I upgraded to the 1TB plan and now my whole catalog is in the cloud. But after hours spent tagging faces in Lightroom Classic, I sadly learned that all those names/keywords are NOT in the Cloud. So now I have to do it all over! 

 

Can you understand why I might be a little grumpy?

Votes

Translate

Translate

Report

Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines