BN HTRd Dataset

The BN-HTRd dataset is a benchmark resource for offline Bangla handwritten text recognition (HTR), focusing on accurate line and word segmentation within document-level images. Current research leverages deep learning models, often incorporating object detection frameworks and techniques like Hough transforms, to improve segmentation and recognition accuracy. This dataset's availability significantly advances research in HTR for low-resource languages like Bangla, impacting both the development of improved algorithms and the accessibility of information in diverse scripts. Furthermore, related datasets are being developed to address challenges in other areas such as biometric authentication using VR headsets and forged document detection.