| --- |
| library_name: onnx |
| tags: |
| - ocr |
| - paddleocr |
| - pp-ocrv6 |
| - text-recognition |
| - onnx |
| - inference4j |
| license: apache-2.0 |
| pipeline_tag: image-to-text |
| --- |
| |
| # PP-OCRv6 Tiny Rec — ONNX |
|
|
| ONNX export of [PaddlePaddle/PP-OCRv6_tiny_rec_onnx](https://huggingface.co/PaddlePaddle/PP-OCRv6_tiny_rec_onnx), the lightweight tiny text recognition model from PaddleOCR PP-OCRv6 (~4.5 MB). Features a compact 6,904-character CTC dictionary. |
|
|
| Mirrored for use with [inference4j](https://github.com/inference4j/inference4j), an inference-only AI library for Java. |
|
|
| ## Original Source |
|
|
| - **Repository:** [PaddlePaddle](https://huggingface.co/PaddlePaddle/PP-OCRv6_tiny_rec_onnx) |
| - **License:** apache-2.0 |
|
|
| ## Usage with inference4j |
|
|
| ```java |
| try (var ocr = OcrPipeline.builder().build()) { |
| List<TextRegion> lines = ocr.read(image); |
| } |
| ``` |
|
|
| ## Model Details |
|
|
| | Property | Value | |
| |----------|-------| |
| | Architecture | PP-OCRv6 Tiny Recognition (SVTR-CTC) | |
| | Task | Text recognition | |
| | Model size | ~4.5 MB | |
| | Vocabulary size | 6,904 characters (+ blank and space) | |
| | Input format | [batch, 3, 48, W] (BGR, height 48, dynamic width) | |
| | Output format | [batch, time, 6906] (CTC softmax probabilities) | |
| | Original framework | PaddlePaddle (PP-OCRv6) | |
| | Technical report | [PP-OCRv6](https://github.com/PaddlePaddle/PaddleOCR) | |
|
|
| ## License |
|
|
| This model is licensed under the [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0). Original model by [PaddlePaddle](https://github.com/PaddlePaddle/PaddleOCR). |
|
|
