Turning likes into action: PetRescue uses facial recognition to turn spotlight on rescue dogs
Dog photos on social media tend to receive tons of love, with followers of these accounts numbering in the thousand and views sometimes in the millions. Rescue dogs on the other hand find similar attention hard to come by as they do not enjoy the privilege of having a human owner to build their social presence.
To help these dogs, PetRescue, an organization that creates innovative programs and delivers them for free to help thousands of rescue pets across Australia, created PetMe, an artificial intelligence (AI) platform capable of analyzing dog photos on Instagram and finding similar dogs available for adoption.
Created with its agency DDB Australia, the organization wanted to use PetMe to match users who follow popular dogs on social media to a candidate of their favorite breed. It does this by identifying breeds, temperament, size and coat of these popular dogs, and matches those qualities to more than 90,000 available rescue dogs nationwide.
Users can also post any dog image to Instagram using a hashtag #PetMe and the AI will recognize the breed and auto-comment back to everyone offering a dog of matching qualities up for adoption.
Alex Newman, a senior art director at DDB, tells The Drum that the task was to get as many dogs adopted as possible and that it first needed to find an audience that was willing to own a dog. The agency did some digging and discovered several articles on the influence of Instagram famous dogs, some of them who had so much of followers that they flourish with e-commerce and merchandise.
As people wanted these dogs so badly, the agency knew they were onto a powerful cultural behavior and Instagram was the perfect place to be.
“From that insight, we thought of different ways we could turn soft information, like the images on Instagram, into hard information, like a keyword that could be input into a database of rescue dogs. Image recognition has become so good these days, it seemed like the best way to convert images into text,” explains Newman.
“Then, we had to get influencers on our side and as most of them have rescue dogs, they were excited to be a part of the project. All they had to do was post their next image with hashtag PetMe.”
Newman says that offering a similar dog up for adoption for all of the popular dogs’ followers and potential owners to see, proved to be a very clever way for PetRescue to talk to its audience because people hate to get interrupted by ads on Instagram.
“Turning the comments section into an ad space was simple and effective. Basically, they didn't have to like or even follow, PetRescue’s page to see the dog,” he explains.
Creating a campaign that relied heavily on AI and machine learning was not without its challenges though. According to Guilherme Machado, a creative copywriter at DDB Australia, Instagram banned bots halfway through development, which meant the agency had to get smart with how it posted.
To circumvent this obstacle, the agency built a natural pause into the bots reply, where instead of responding to the use of the hashtag immediately, we waited a minute or two before commenting back.
In addition, the agency also cycled between bots making it seem like different people were replying to the posts. These modifications allowed the posts to slip under Instagram’s radar of bot security.
“Another challenge was the number of breeds that exists and their characteristic. We created a database of more than 200 breeds with their characteristics such as size, temperament, intelligence and so on,” explains Machado.
“This is what fuel the machine learning aspect of the project. The system was able to make connections between pure breeds and mutts, allowing it to find the next best dog when no exact breed match was found in a shelter.”
PetMe also had several trial and error phases, which means there were a lot of moving digital parts that had to work in unison. “From our end, the AI, its machine learning database and the bots for responding had to all be working as one,” explains Machado.
“On the clients’ end, their database had to be groomed and updated to be compatible. The moral of this project was that the ideas of this complexity are constant problem-solving exercises.”
The agency claims PetMe shed a light on PetRescue, increasing the brand exposure and raised the number of adoptions to 14,000 in the last six months. Some NGOs around the world, such as Brazil and the US, have also shown interest in “adopting” the PetMe platform to their markets.