A large brick and mortar retail chain came to us seeking help with optimizing and monetizing their data capture from photos of customer receipts
KUNGFU built a hybrid model, leveraging client’s existing cloud APIs, a custom YOLOv3 boundingbox detection algorithm, and a Generative Adversarial Network (GAN) model to enhance and extract appropriate textual data from receipts.
We were able to meet requirements for latency of <2 seconds and achieve an average F1 Score of 0.88 for detected fields.
A large brick and mortar retail chain came to us seeking help with optimizing and monetizing their data capture from photos of customer receipts
KUNGFU built a hybrid model, leveraging client’s existing cloud APIs, a custom YOLOv3 boundingbox detection algorithm, and a Generative Adversarial Network (GAN) model to enhance and extract appropriate textual data from receipts.
We were able to meet requirements for latency of <2 seconds and achieve an average F1 Score of 0.88 for detected fields.
