Starstruk Logo MarkStarstruk

v3.7.0

dgtlmoon/changedetection.iov3.7.0Jun 11, 2026by Bobholamovic

AI Summary

Version 3.7.0 introduces the PP-OCRv6 engine, delivering significant accuracy improvements and unified support for 50 languages. The release features a three-tier model architecture optimized for different hardware environments and includes specialized recognition capabilities for complex industrial scenarios.

Key Highlights

  • Release of PP-OCRv6 with superior accuracy surpassing mainstream VLMs
  • Unified model covering Chinese, English, Japanese, and 46 Latin-script languages
  • Specialized improvements for digital displays, dot-matrix characters, and industrial text
  • Faster inference with 5.2x CPU speedup (OpenVINO) and 6.1x Apple M4 acceleration
  • Three-tier model architecture (tiny, small, medium) for edge, mobile, and server deployment

New Features

  • PP-OCRv6 accuracy boost (+4.6% detection, +5.1% recognition)
  • 50-language unified OCR support
  • Enhanced industrial and specialized text recognition
  • Hardware-optimized inference performance
  • Scalable model sizes for various device constraints

Full Release Notes

## 2026.6.11 v3.7.0 released

* **Release PP-OCRv6**

    - **Accuracy boost**: Medium tier achieves +4.6% detection and +5.1% recognition over PP-OCRv5_server, surpassing mainstream VLMs (Qwen3-VL-235B, GPT-5.5) with only 34.5M parameters.
    - **50 languages unified**: Single model covers Chinese, English, Japanese, and 46 Latin-script languages — no model switching needed.
    - **Specialized scenarios**: Major improvements in digital displays, dot-matrix characters, tire prints, and industrial text recognition.
    - **Faster inference**: 5.2× CPU speedup (OpenVINO), 6.1× on Apple M4 (tiny), 0.13s on A100 GPU.
    - **Three tiers for all scenarios**: tiny (1.5M) / small (7.7M) / medium (34.5M) for edge, mobile, and server deployment.

## 2026.6.11 v3.7.0 发布

* **发布 PP-OCRv6**

    - **精度全面提升**：medium 档相比 PP-OCRv5_server 检测精度提升 4.6%、识别精度提升 5.1%，以仅 34.5M 参数超越 Qwen3-VL-235B、GPT-5.5 等主流视觉语言大模型。
    - **50 种语言统一支持**：单一模型覆盖中文、英文、日文及 46 种拉丁语系语言，无需为不同语种切换模型。
    - **专业场景增强**：数码显示屏、点阵字符、轮胎印字、工业字符等传统 VLM 难以覆盖的场景识别能力大幅提升。
    - **推理速度更快**：medium 档 CPU OpenVINO 推理加速 5.2×，tiny 档 Apple M4 加速 6.1×，A100 上仅需 0.13s。
    - **三档模型覆盖全场景**：tiny（1.5M）/ small（7.7M）/ medium（34.5M）分别面向端侧/移动端/服务端部署。

**Full Changelog**: https://github.com/PaddlePaddle/PaddleOCR/compare/v3.6.0...v3.7.0