en_PP-OCRv5_mobile_rec

en_PP-OCRv5_mobile_rec is an image-to-text model for English text line recognition, optimized for mobile deployment. It is a specialist model hosted on gigarouter as a managed, OpenAI-compatible API.

Key Strengths

• High accuracy on English text lines: achieves 85.3% accuracy under strict per‑character evaluation (any incorrect character marks the entire line as wrong).
• Lightweight architecture designed for efficient inference on resource‑constrained devices.
• Part of the PP-OCRv5 family, developed by the PaddleOCR team.

Benchmark Results

Accuracy is measured with strict line‑level scoring: a line is considered correct only if every character (including punctuation) is recognized correctly.

Example Output

Given an input image, the model returns the recognized text and a confidence score. Below is a sample visualization of a correctly recognized line:

Best For

• English text recognition in mobile or edge applications where model size and speed are critical.
• Use as a standalone text line recognizer or integrated into a full OCR pipeline (including document orientation, text detection, and text line orientation).

Additional Information

For further details, refer to the PaddleOCR documentation and the PaddleOCR repository.