carsales.com has successfully developed and deployed its Cyclops Image Recognition Software to improve the efficiency, accuracy, and consistency of Australia’s leading online automotive classifieds business.
The cutting-edge software, which the company claims has a 97.2% accuracy rate, automatically selects and assigns angles to each image uploaded by the company’s photographers directly onto the carsales platform to help showcase various aspects of the vehicles on sale. Cyclops also has an AI feedback mechanism so it automatically learns from any mistakes it makes thus improving its performance rate over time.
Cyclops was conceived during one of carsales’ biannual Hackathons in November 2016. The hackathon is a three-day forum where the company’s product and technology teams work alongside employees from other departments in small teams to think laterally and generate technical, creative and innovative ideas.
According to Michael Ridgway, carsales’ Director of AI Research, the Hackathon has been a valuable initiative in more ways than one, “Hackathons are a fantastic opportunity for all our staff to put their heads together and come up with ideas that help carsales continue to innovate and evolve to provide users with the best possible online buying and selling experience.”
{loadposition ray}
The theme for the Hackathon was ‘cognitive computing’, with the Cyclops team taking out the International Award for its ability to efficiently categorise vehicle imagery regardless of country, within the media library systems which integrate with carsales’ international acquisitions.
Agustinus Nalwan, who invented the idea and lead the winning team, said; “Prior to the implementation of Cyclops on the carsales platform, every image of a car uploaded into our carsales Media Library by Photo Services staff was manually categorised according to the angle featured in the image. Now with Cyclops, each image is automatically assigned an angle.”
Over the next 2 months, the AI team will further integrate the technology for photos uploaded by users and will notify users to upload additional images if necessary.
“For example, a user who is selling an SUV and has uploaded images of the front exterior, side exterior, front seats, and dashboard will be prompted to upload an image of the backseat too. This is because Cyclops knows this is one of the most common features people research when purchasing an SUV,” said Nalwan.
The technology is being rolled out globally across carsales’ international businesses and over the next few months the AI team will further develop the Cyclops technology to create an image comparison tool to provide users with the ability to visually compare the backseat photos of say a Honda CRV, Mazda CX-5 and Ford Kuga.
“The beauty of AI means that Cyclops will continue to develop and be adapted to deliver many more customer benefits in the future,” said Ridgway.
Image courtesy http://sbiii.com/cyclops/cyclops.html