AustraliaTechnology and telecom2016 – ongoingCarsales
Cyclops vehicle-photo classification
Carsales said Cyclops removed 9,000 hours of annual photo-tagging work
Problem
Carsales photographers manually classified vehicle images for its image library. The company put that work at 9,000 hours annually.
Approach
An internal team developed Cyclops after a 2016 hackathon, using stored vehicle images for training. It deployed the classifier as a microservice to tag photographed views. Carsales then applied the model to customer uploads so buyers could compare corresponding vehicle views.
Results
Takeaway
A classifier built for staff also supported customer-facing photo comparisons. Its accuracy improved after early problems with training data and misclassification.
Categories
- Industry
- Technology and telecom
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Computer vision
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
CIO reported Jason Blackman's account of developing and deploying Cyclops. He attributed the removal of annual tagging work and accuracy of up to 98% to the system.
Full details are in the original sources.
