PP-LCNet_x1_0_textline_ori
PaddlePaddle/PP-LCNet_x1_0_textline_ori
A popular open image-to-text model, with 274.6K downloads a month. gigarouter benchmarks and hosts it as an OpenAI-compatible API.
about this model
This model performs text line orientation classification, a specialized image-to-text task that identifies whether a text line is oriented at 0 degrees or 180 degrees. It is designed to correct rotated text lines before they enter an OCR pipeline, improving overall recognition accuracy in document scanning, license/certificate photography, and similar workflows.
Key Strengths
- High accuracy: 98.85% recognition average accuracy on the intended task.
- Compact footprint: model storage size is only 0.96 MB.
- Two-class output (0 degrees and 180 degrees) enables straightforward post-processing correction.
- Based on PP-LCNet_x0_25, a lightweight architecture optimized for efficient inference.
Best For
Preprocessing text lines in OCR pipelines where capture device rotation may produce misoriented text. The model is particularly useful for batch document processing, identity document verification, and any scenario requiring robust text orientation correction prior to recognition.
Performance
| Model | Recognition Avg Accuracy (%) | Model Storage Size (M) | Introduction |
|---|---|---|---|
| PP-LCNet_x1_0_textline_ori | 98.85 | 0.96 | Text line classification model based on PP-LCNet_x0_25, with two classes: 0 degrees and 180 degrees |
When provided with a text line image, the model outputs a class ID and confidence score. The example below shows a sample image classified as 180 degrees with a score of 0.99829.
Links
We're benchmarking and onboarding PP-LCNet_x1_0_textline_ori as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.