AI Protects Pups This Christmas: Ensuring Dogs Are for Life

As the holiday season brings a surge in demand for furry companions, concerns rise over the practices of puppy farmers exploiting this trend.

Puppy farmers try to appeal to holiday shoppers wanting to bring a pup into their home by enticing them with festive ads online. However, the pups displayed in these ads are often sick and may not live past January, a sad reminder of the famous slogan, “A dog is for life, not just for Christmas” from U.K. charity Dogs Trust.

To combat this alarming trend, one pet rehoming site has enlisted AI to ensure that puppy farmers and animal abusers do not get the chance to exploit the holiday’s demand for dogs.

U.K.-based Pets4Homes is using an AI system called Spotty to detect and automatically block any adverts that explicitly link animals to being gifts. Using image recognition, the system can detect an ad that features a puppy next to a Christmas tree, making it appear as a present.

Spotty analyzes images for indicators of health and well-being, such as the animal's physical appearance and signs of stress or neglect, with the system able to identify environmental factors like cramped spaces or unsanitary conditions.

Spotty uses convolutional neural networks for image analysis, having been trained on a proprietary dataset of over three million images, annotated details about health, breed and environmental conditions.

Pets4Homes told AI Business that they review and annotate close to 7,000 images per day and are constantly feedbacking data for retraining Spotty.

Here are just a couple of the ads that Spotty has taken down:

Credit: Pets4Homes

Could it be expanded beyond cats and dogs? The system's architecture allows for the potential inclusion of additional species in the future, should the need arise.

Looking Beyond Cute Puppy Pics

It is not just about spotting issues in images of the pets themselves - Pets4Homes enlists the help of OCR (Optical Character Recognition) technology to check the authenticity of documents covering puppy vaccinations, Kennel Club certificates and DNA testing results. Using the open source solution Tesseract OCR, Pets4Homes scans more than 2,000 documents a day, with extracted text data then cross-referenced with its internal database for verification.

For monitoring conversations between breeders and buyers, Pets4Homes uses simple natural language processing (NLP) techniques to flag potentially fraudulent or unethical interactions. Chat messages are analyzed, with sentiment analysis and keyword detection employed while custom algorithms look for patterns like repetitive messaging or uses of flagged terms.

Pets4Homes even looks out for financial fraud, employing a mix of statistical models and machine learning to identify unusual financial patterns, like sudden high-volume transactions or unusual payment distributions across multiple accounts. The system also monitors user account behavior, flagging accounts with rapid transactional activities across different breeds as potential risks.

There is even some anomaly detection in listing details – using cluster analysis and outlier detection algorithms to identify listings that deviate from normal patterns, such as unrealistic pricing or inconsistent breeder information. Pets4Homes said that due to large swings in price levels due to the demand for pets before, during and after the COVID lockdown, it had to dynamically update its algorithm to adjust for market trends.

Pets4Homes would not disclose how much it spends on all this tech, but the rehoming platform did reveal to AI Business that its primary expenses include cloud computing resources for training and deploying the AI models, ongoing data acquisition for model refinement and a dedicated team for system development and maintenance.

More to Come

This year marks the second time Pets4Homes will use its tech arsenal for Christmas. Last year, it blocked 2,300 misleading advertisements. So far in 2023, the platform has seen a 70% drop in Christmas-themed puppy adverts.

Axel Lagercrantz, CEO of Pets4Homes told AI Business that away from the holiday season, the tech is used to spot dogs that have cropped ears, an outlawed practice in the U.K., where a dog’s ears are surgically removed purely for looks.

Pets4Homes’ expansion of the system will also see it used to detect banned breeds. The U.K. government announced in October that the XL Bully breed would be banned under the Dangerous Dogs Act.

“We are continually working to enhance the precision of our image recognition models, particularly in identifying subtle indicators of animal health and well-being,” Lagercrantz said.

The CEO said the tech itself is also expanding to include improvements to its database to cover a wider array of breeds and health conditions. Also getting an upgrade is Pets4Homes’ NLP models in order to obtain more nuanced chat analysis.

This article first appeared on IoT World Today's sister site, AI Business.