Summary

Reverse image search engines provide varying results in identifying the photographer and source of portrait images, with mixed outcomes on crediting creators and facial recognition capabilities.

Abstract

The article explores the effectiveness of AI reverse image search engines, particularly when it comes to portrait images of people's faces. It highlights that while these engines can find similar images and provide extensive data on image usage, they often fail to credit the original photographer or link to the source image. Tests conducted with Google, TinEye, and Pinterest reveal that none of these popular search engines offer facial recognition or identify individuals. Google's search engine returns numerous links but lacks the ability to identify the original source or photographer. TinEye provides valuable data on image popularity and credits the photographer but does not display similar images. Pinterest offers a visual array of similar images and ads but also falls short in crediting the original creator. The article underscores the importance of image credit for creators and the potential for misattribution, as well as the privacy implications of facial recognition technology.

Opinions

The author believes that photographers and creators should be credited upfront in search results, especially for popular free-to-use images.
The article suggests that Google's AI has a shortcoming in not identifying the original upload link or the photographer's name.
TinEye's approach to returning search data is considered extremely valuable to creators for its quick stats on image popularity and reference to the photographer and source link.
The author expresses a preference for Pinterest's search results, appreciating the visual clues and the integration of ads, but also points out the lack of photographer and source file referencing as a disappointment.
There is concern over the loss of control creators have over their images once they are released under a common license, including the possibility of other photographers claiming authorship.
The author has reservations about facial recognition technology, noting its controversial nature and the privacy concerns it raises.
The article advises caution when uploading portrait images online, emphasizing the benefits of reverse image searches for verifying the authenticity of profiles on social or dating apps.
Regular reverse image searches are recommended for creators to monitor unauthorized use of their images.
The author concludes that the ideal outcome of a reverse image search is to transmit data about the creator's original upload link and a description of the image, without engaging in facial recognition.

Entry screen on Pinterest of a similar image search, based on a popular image of a young woman smiling. Original uploaded photo by Michael Dam on Unsplash.

Every face tells a story in a reverse image search, but outcomes vary depending on the search engine

How do AI reverse image portrait searches hold up?

Reverse image searches are defined as, in essence, Searching for similar images. An image of choice is submitted to the AI search engine, which in turn scans the entire internet for either the same image and its source. The search engine also searches for similar images based on visual clues such as colors, subject, landscape or building landmarks, camera angle, light sources, depth of field, or even blurriness or out of focus areas.

AI reverse image search engines can produce impressive results. They reference not only dozens, if not hundreds of similar images, they can also source the location of the original image (if posted online as publicly-accessible image), availability of different image resolution, descriptions of the photo subject, and in some instance, references to location and maps.

Read more on What is a reverse image search? here.[1]

AI search results can get a bit trickier when we search for similar images of a person’s face. Contrary to first beliefs, popular reverse image search engines, such as Google’s, TinEye’s, or Pinterest’s, do not render facial recognition searches, and do not identify individuals.

When I did test runs with these three search engines, Google, TinEye, and Pinterest, I was keen to find out these three pieces of information:

1. Is the name of the photographer of the source image credited in the search?

2. Are similar artistic images showing up in the results?

3. How many commercial links are referenced in the results?

Google’s search engine

I’ve uploaded an immensely popular portrait of an exuberant young woman by Michael Dam, sourced on Unsplash, as free to use under the Unsplash License, to the Google reverse image search engine. I begin my reverse image search.

Screenshot of Google’s reverse image search results. Uploaded photo by Michael Dam on Unsplash.

Not surprisingly, Google’s search engine returns dozens of links of where the image was used.

However, Google’s AI is not able to identify the location of the source original upload link, nor is the photographer’s name. That’s a shortcoming. I’ve sent feedback to Google.

Especially for free-of-use popular images, but also in general, photographers and creators of an image need to be credited in the search results upfront.

While most of Google’s results seem legit links to websites where the image is used, there are several click-bait links to sites that go nowhere. Several of them are to LinkedIn to random users, others link to a Facebook internal page that goes nowhere.

TinEye’s reverse search results

TinEye’s result page comes in differently. TinEye does not return visual replicas of the source image, but rather, it renders a display of a number of findings (2,279 results) and the photographer’s info and original source.[2]

TinEye’s search engine indexed number data and first upload reference and photographer’s link.

TinEye’s search data is extremely valuable to the creator, as quick stats about the popularity of an image are returned. For a user, the reference to the photographer and the source link is valuable in case more information is needed on the image.

TinEye does not return similar images.

Pinterest reverse search results

Pinterest’s search engine engages with the first click on a Pinterest-sourced image automatically. As soon as an image is clicked, the page fills with versions of the source image, similar images, and a handful of ads geared towards the audience of the sourced image.

Pinterest has an embedded search engine that launches on every image click. Similar images are displayed at once, along with handful of ads topically related to the image search.

I like Pinterest’s results a lot, as they present a smorgasbord of visual clues, ads notwithstanding.

Only when activating the image’s reverse search button, see magenta arrow, does Pinterest engage in the search for this image.

Pinterest’s activated reverse image search reveals the popularity of the image. Ads are more subtly integrated into the image, and not upfront. The source image and photographer was not found.

Because the image referenced is such a popular image used thousands of times, the same image shows up over and over again.

As in Google’s reverse image search, the photographer is not referenced, nor is the original source file. That’s really disappointing, and a shortfall of this search.

One search result even links back to another photographer on Unsplash, claiming authorship of the image as hers. If I were the creator of this image, it would be very upsetting to me.

Research findings on doing quick reverse image searches

Popular image search engines do not engage in facial recognition results, and do not identify the person.

Creators are, in most searches, not credited, nor is their original source upload.

For creators, uploading a beautiful and relevant image free of charge under the common license will also mean that control over the image is lost, even to the extent that other photographers claim it as theirs.

Why do an actual facial recognition search?

To conduct actual facial recognition searches means that the software in use is mapping a face based on biometric data points. The software then stores a faceprint of such mapping which is then used to compare other databases of faces, and presumably, the stats that come with them.

It’s still largely unknown how often facial-recognition technology is being used and where in the United States [3].

Facial recognition software is not fully accurate, and highly controversial.

Personally, I get a bit of knot in my stomach thinking of the possibility of my face indexed and mapped in a database somewhere. So, I think it probably to be best to be utterly cautious when uploading a portrait image of myself online.

The benefits of doing a similar image search of a person’s face

If we are ready to befriend someone on a social dating app, but not 100% sure about that person, we can do a reverse image search and see what pops up.

Some face reverse image searches don’t render many results. Most searches might focus on accessories, such as clothes or jewelry, and not on the face, to render results.

Searches based on personal uploaded photos which have not been already uploaded on the web, return mostly random shots, and nothing worthy of any real data.

What to look out for

For creators, doing reverse image searches from time to time might become crucial when looking for unauthorized usages of images.

Uploading one’s portrait pictures should always be done with caution. While social platform sites like IG or FB protect their content, others do not.

Facial searches come with a caveat for the average user. The results are most likely non-revelatory, meaning that the only data that can be transmitted is that of the creator’s original upload link, a reference to the creator, and a description of what an image is about.

And that’s a really good thing.

References/links:

[1] Read up about reverse image searches: https://evaschicker.medium.com/what-is-a-reverse-image-search-70c272fef087

[2] TinEye search-by-image engine: https://tineye.com/

[3] https://www.cnn.com/2022/08/05/tech/facial-recognition-bans-reversed/index.html

Thank you.