Unmodified copies of the six PP-OCRv6 ONNX models published by RapidAI / RapidOCR, mirrored by neavo. The model family originates from PaddlePaddle / PaddleOCR.
本仓库转存 RapidAI 发布的 PP-OCRv6 ONNX 系列,包括文本检测(det)与文本识别(rec),各有 tiny、small、medium 三档。六份权重逐字节保留,未重新导出、量化、裁剪或改写模型元数据。
7d0781614ca1a83d5ad9603f713acb2e74855d72.onnx/PP-OCRv6/; this mirror retains the relative det/ and rec/ paths and original filenames.| File | Bytes | SHA-256 |
|---|---|---|
det/PP-OCRv6_det_medium.onnx |
62,119,454 | 92078b7355007ccfffcd4c8cd441a3afd4538904d06881b29a155e1e679907c2 |
det/PP-OCRv6_det_small.onnx |
9,929,594 | 090f04abcd9d9a7498bc4ebf677e4cb9bdce1fe4197ddb7e529f1ef44e1ff94f |
det/PP-OCRv6_det_tiny.onnx |
1,829,618 | f42c0fbd294d95eac1a550e131b277dac97462c8025fa4b6c3cec1b7894bd3d5 |
rec/PP-OCRv6_rec_medium.onnx |
76,629,984 | eef444829dbbe18d7fea59a3f6eb75647518d2b3a9568d27c92e42940204894b |
rec/PP-OCRv6_rec_small.onnx |
21,234,383 | 6f327246b50388f3c176ae304bd95767ea6dc0c9ae92153ef8cbe210b3c14884 |
rec/PP-OCRv6_rec_tiny.onnx |
4,489,813 | e16e242de5937ad92609223f19bc2aff3727ee40b095f996907c24749bad251b |
Use these models with the corresponding RapidOCR / PaddleOCR preprocessing, recognition decoding and character dictionary configuration. A detection model and recognition model have different responsibilities; they are not interchangeable. Follow the upstream project for the required runtime and pipeline settings.
This repository is a file mirror, not a new training or conversion release and not a Transformers pipeline checkpoint. Downloaded files were verified against the pinned upstream SHA-256 values. SHA256SUMS also covers accompanying files.
No additional inference, accuracy, or GPU compatibility benchmark was performed for this mirror. A .onnx extension alone does not guarantee WebGPU operator coverage.
The upstream ModelScope repository declares Apache License 2.0. LICENSE preserves the Apache-2.0 text and RapidOCR Authors copyright notice supplied by the RapidOCR project. Model family credit belongs to PaddlePaddle / PaddleOCR; ONNX distribution credit belongs to RapidAI / RapidOCR. This mirror preserves those attributions and does not claim authorship of the models.
rec/PP-OCRv6_rec_medium.onnx 77 MBdet/PP-OCRv6_det_medium.onnx 62 MBrec/PP-OCRv6_rec_small.onnx 21 MBdet/PP-OCRv6_det_small.onnx 10 MBrec/PP-OCRv6_rec_tiny.onnx 4 MBdet/PP-OCRv6_det_tiny.onnx 2 MB