PP-LCNet_x1_0_doc_ori

PP-LCNet_x1_0_doc_ori is a document image orientation classification model that identifies the orientation of document images across four categories: 0°, 90°, 180°, and 270°.

Key Strengths

This model is designed to pre-process document images—such as scanned documents or ID photos—that may have been captured at arbitrary rotations. By classifying orientation before OCR, it improves downstream text recognition accuracy. It achieves 99.06% recognition average accuracy with a storage size of only 7 MB.

Best For

• Correcting rotated document images in scanning or photo-capture workflows
• Preprocessing inputs for OCR pipelines to boost overall recognition accuracy
• Digital document management and automated document image quality improvement

Performance

Model	Recognition Avg Accuracy (%)	Model Storage Size (M)	Categories
PP-LCNet_x1_0_doc_ori	99.06	7	0°, 90°, 180°, 270°

Example Output

For a sample input image, the model predicts a class ID of 2 (180° rotation) with a confidence score of 0.90971:

Integration

Gigarouter hosts this model as a managed, OpenAI-compatible API. You can call it directly without any local installation or dependency management. The model can also be used as part of a document preprocessing pipeline that combines orientation classification with geometric distortion correction for more robust document image handling.