Over the past few years, Facebook's made significant strides with their image recognition AI, helping to power a more advanced version of their on-platform image search tool - which probably hasn't received as much attention as it deserves.
There are a heap of ways you can use image recognition to advance your social marketing strategy - for example, if you can identify people who regularly share images of products similar or related to your own on Facebook, and within your local region, that could present an opportunity for you to reach out and offer a discount or special offer to them in the hopes of establishing a connection and getting them to share the word about your business with their network.
It's not as simple a process as blanketing groups with ads, but getting a selected users to post a personal update is more conducive to viral sharing, and targeting those who are already sharing related images could be a great way to facilitate such.
With Facebook's search capacity, this is already possible, and this week, The Social Network has shared a new research paper which underlines just how advanced and focused their image search tools have already become.
First off, back in February Facebook released a research update on their advance image recognition AI and how it was now able to identify more actions and objects within photos, which have been built into their on-platform search function.
As you can see in these examples, you can now search for places or events and Facebook will be able to identify such, even if the image wasn't tagged with that information.
Facebook's latest research paper details how they've advanced this capacity even further by using their Unicorn indexing system to provide more contextually relevant results.
Unicorn was originally built to power Facebook's Graph Search, which was their advanced search tool that uncovered a heap of data and insights - so much data, in fact, that Facebook had to scale it back to protect user privacy and maintain control over their insights.
In this new iteration, Facebook's using a more contained version of Unicorn to provide better image matches, based on relevant context, personal matches and public posts.
The explanation of the back-end workings of the system is fairly technical, but in essence, Facebook's new systems now have a better understanding of both the text and image content of each post, which helps the system provide more contextual matches, enabling users to find what they're after within the billions of photos being uploaded to The Social Network every day.
And as noted, that can be hugely beneficial.
For example, if you run a bike store, you can check out all public posts within a city or region, and within a certain date range, that include bikes.
This could show you common places where people are taking photos, which might help provide a better understanding of where such groups are hanging out, or, as noted, you may be able to identify users who regularly post about bikes and bike wear in your region, and who generate a lot of engagement, helping to identify potential influencers to collaborate with. This has always been possible to a degree, but Facebook's advanced image recognition systems now drag in more relevant results, related to more specific queries, which could open up new opportunities for those who go looking.
As noted, viral sharing is significantly influenced by individuals within groups, as opposed to broad stream blasts, as demonstrated in this graph from Derek Thompson's book 'Hit Makers: How Things Become Popular'.
The best option, Thompson says, is the final one here, diffuse broadcast, where you're utilizing both traditional broadcast methods and individual word-of-mouth to help spread the word. In line with this, identifying Facebook micro-influencers in this way, while also targeting a broader group with paid ads, could be a great way to expand your presence in a localized market.
As noted, it takes a little more work to conduct the research and get in touch with relevant influencers, but Facebook's advanced image recognition tools make this easier, with more options on how you find the right users based on their posted image content.
This is just one example of how Facebook's advanced image search capacity could provide benefit. Worth taking a look to see what you can find, and how Facebook image search is evolving.