As a contributor myself I have to confess something, I really consider technology a part of our daily lives. Not only that they take care of certain repetitive tasks, but they are able to approach topics in a totally different manner. In order to make things even more interesting and complicated I aimed myself to find answers to some of the questions that puzzled me till today:
- What is lying behind the submission and review process?
- Who is reviewing the content, human or AI?
- How can AI help ease submission and review process?
In order to find answers to the above questions I decided to contact the biggest stock photo agencies:
To be honest before writing this article I was a little bit skeptical. My initial thought was that nobody will bother to answer me. Things turned out to be quite the opposite. Luckily I’ve got quite a detailed answer from dreamstime and fotolia. I want to thank them for the quick and detailed answer which I would like to share with the contributor community.
What is lying behind the submission and review process
Regarding the submission and review process most of my initial feeling has been confirmed. Dreamstime is using technology to their favor. What does that mean?
They have integrated artificial intelligence in a lot of repetitive tasks. Please find their official answer regarding my questions:
Dreamstime
We don’t want repetitive, common tasks to be done by humans, we want them to the creative part and to fine tune the collection this way. The AI is also employed on the search engine’s side and that fine tunes the results too. There are hundreds of instances where the AI engine works, from submission process (where we suggest categories), through review (where we validate some of the things inside the image, such as copyright, content filter, model release and trigger various alerts, prioritizing review tasks) all the way to download, where the user can see various images more suitable for his/her user experience on Dreamstime.
The Adobe stock team generally don’t share detailed insights in the technology they are using, but for this questions they did not hesitate to answer.
Adobe Stock
Usually we do not disclose any information about our moderation, but I can answer those questions for you:
In their case, without giving much detail, the situation is the following. As like dreamstime they have AI technology implemented in the submission process.
Adobe Stock
AI is used only to help a bit on that huge workflow our moderators have to deal with: for example, it automatically rejects images that are very out of focus or images with people on them but without a model release are put automatically on remind to complete.
Regarding the review process, this is being performed by people.
Who is reviewing the content, human or AI?
Regarding the review process that statement of Dreamstime is clear.
All images are reviewed by humans, but some significant tasks are AI-based.
If this is true and I think that they don’t have any interest to lie then I have to admit that what these guys/girls are doing is outstanding. Just to give you some insight in the numbers:
- current database: 122 million stock photos
- registered contributors: 28,437,242 users out of which: 583,024 photographers
- 3,949,503 monthly images
Going through roughly 4 million images every month is something hard to comprehend.
Out of focus or excessive noise
As some of you may have noticed lately a lot of people are complaining about Shutterstock about that fact that their images are being rejected because of focus or noise issues. To be honest it is really to state that the reason is due to human reviewer or some sort of pseudo error which AI is generating. Something is causing the increased number of rejections.
It is very possible that AI is making a pre evaluation before human reviewer is making the final decision.
How AI can help ease submission and review process
We have to admit that artificial intelligence is becoming more and more part of our lives. Even if we don’t notice it every time AI is performing a lot of repetitive and boring tasks.
With this in mind it becomes more obvious that this is not different in case of stock content creation and the whole reviewing process. Machines are designed to perform repetitive tasks. Even though it is unlikely that they are unable to replace human reviewers soon (let’s hope), they can ease the entire process. Let’s see some examples where artificial intelligence is showing its presence:
- uploading process
- batch editing
- automatic category assignment
- verification of focus / noise in an image
- keyword suggestion based on content
- verification of image/footage fulfill minimum requirements (e.g. size, bit rate, image format, etc)
Some of you might say that some of the above processes can be achieved in a programmatic way (e.g. writing a small piece of code to check certain tasks) . If the task is being performed based on content it is no longer a simple piece of code like adding 2 numbers. Indeed for example verification if the image is in focus or has excessive noise can be considered an implementation of artificial intelligence. Additionally if you add the possibility for the software to learn at each iteration then this becomes obvious.
While not all major stock content agencies are embracing technology and implementing from tomorrow they have noticed that they can have some benefits.
Conclusion
Whether we like it or not artificial intelligence is becoming an important part of our life.
While we can’t be 100% sure what is happening behind the scenes, most stock photography agencies have admitted the integration of AI in their evaluation process.
Sometimes this can be beneficial for us as it eases the entire process. As a consequence we either finish faster this task or use the extra time to create more content.
But sometimes it an lead to some frustration if the piece of code is not optimized well. As a consequence it can lead to image rejections, lost time or frustration.
Given these points it is obvious that artificial intelligence is becoming more and more efficient and we the future path is quite straight forward.
