Counterfeit shoes detection model
For a stealth mode startup in U.S. we created an image classification model for identifying real Vs counterfeit shoes by using close up images of shoes. The model assessed shoe quality, material and other nuances and provides 97% accuracy.
Business Impact
Achieved 97% accuracy in identifying fake shoes.
Helping curb counterfeit market in high end shoes industry.
Methodology
Created image classification model for identifying Real Vs Counterfeit shoes. This is deployed as a service on a website, where the customers can upload close up images of their shoes. The model uses these images and analyses quality, material, finishing and other nuances to distinguish between real and counterfeit shoes. Model uses a ResNet-152 network as a backbone, with customer top layers.