Inspiration
Newer employees who work for big retail companies (such as Nike, Addidas etc struggle to identify their older models, due to an ever expanding product base)
What it does
retAI-L is a platform that assists employees in discovering products that customers once had and can be done through images and not just bar codes
How we built it
Using YOLOV5 model we did transfer learning on a set of shoes so that it can classify the images into the product catalog
Challenges we ran into
Small dataset of shoes (50-100), shoes were highly similar, thus more colour decoding might be needed to improve on the accuracy
Accomplishments that we're proud of
- Developing a model within the span of 1 day
- Accuracy was around 90% ## What we learned
- Object identification may be prone to errors if the products adopt the exact same shape, thus colour gradients have to be highly differentiated to distinguish the products ## What's next for retAI-L
- allow for better colour filtering in data preprocessing for the trained models
- retrieve more images of the same shoe
- expand into more shoes
Log in or sign up for Devpost to join the conversation.