Characters Detection on Namecard with faster RCNN

TL;DR

This paper applies Faster R-CNN to detect characters on namecards, addressing data scarcity and class imbalance through data augmentation and a 'fake' data generator, achieving high accuracy and improved consistency.

Contribution

The paper introduces a data augmentation and 'fake' data generation method to enhance Faster R-CNN performance on small, imbalanced datasets for namecard character detection.

Findings

Achieved over 80% IoU without data augmentation.

Achieved 80% mAP score with Faster R-CNN.

Data augmentation reduced variance in mAP.

Abstract

We apply Faster R-CNN to the detection of characters in namecard, in order to solve the problem of a small amount of data and the inbalance between different class, we designed the data augmentation and the 'fake' data generalizer to generate more data for the training of network. Without using data augmentation, the average IoU in correct samples could be no less than 80% and the mAP result of 80% was also achieved with Faster R-CNN. By applying the data augmentation, the variance of mAP is decreased and both of the IoU and mAP score has increased a little.