Seeing Straight: Document Orientation Detection for Efficient OCR
This addresses a critical pre-processing step for OCR in real-world settings, particularly for scanned or photographed documents, but is incremental as it builds on existing vision models.
The paper tackles the problem of document orientation detection to improve OCR performance, introducing a new benchmark (ORB) and a lightweight classification pipeline that achieves 96% and 92% accuracy on English and Indic datasets, boosting OCR models by up to 14% for closed-source and 4x for open-weights models.
Despite significant advances in document understanding, determining the correct orientation of scanned or photographed documents remains a critical pre-processing step in the real world settings. Accurate rotation correction is essential for enhancing the performance of downstream tasks such as Optical Character Recognition (OCR) where misalignment commonly arises due to user errors, particularly incorrect base orientations of the camera during capture. In this study, we first introduce OCR-Rotation-Bench (ORB), a new benchmark for evaluating OCR robustness to image rotations, comprising (i) ORB-En, built from rotation-transformed structured and free-form English OCR datasets, and (ii) ORB-Indic, a novel multilingual set spanning 11 Indic mid to low-resource languages. We also present a fast, robust and lightweight rotation classification pipeline built on the vision encoder of Phi-3.5-Vision model with dynamic image cropping, fine-tuned specifically for 4-class rotation task in a standalone fashion. Our method achieves near-perfect 96% and 92% accuracy on identifying the rotations respectively on both the datasets. Beyond classification, we demonstrate the critical role of our module in boosting OCR performance: closed-source (up to 14%) and open-weights models (up to 4x) in the simulated real-world setting.