Paddle OCR v2.10.0 not support 【DB++】，Why？

[wujun]

🔎 Search before asking

• I have searched the PaddleOCR Docs and found no similar bug report.
• I have searched the PaddleOCR Issues and found no similar bug report.
• I have searched the PaddleOCR Discussions and found no similar bug report.

🐛 Bug (问题描述)

paddleocr.py noly suport 【DB】 in SUPPORT_DET_MODEL，other algorihm like ‘DB++’ not supported， Why

🏃‍♂️ Environment (运行环境)

'''bash
OS Windows 10
Env venv
Python 3.9.0
PaddleOCR 2.10.0
RAM 64.00 GB
CPU Intel i9-1300K
CUDA 12.6
'''

🌰 Minimal Reproducible Example (最小可复现问题的Demo)

[GreatV]

PaddleOCR v2.10.0 does not currently support the “DB++” detection algorithm in its official paddleocr.py pipeline (i.e., via SUPPORT_DET_MODEL), and this is by design based on the current project implementation status. There are three main reasons for this:

1. DB++ is still under experimental or internal evaluation status:
   Although DB++ may have been proposed in research or internal code branches, it has not been officially registered or integrated into paddleocr.py’s supported model list (SUPPORT_DET_MODEL at https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.10/tools/infer/predict_det.py or related files). Only “DB” is listed as a supported detector in v2.10.0.

2. Lack of configuration and pretrained models:
   There is no official configuration file (in configs/det/) or pretrained inference model released for DB++ in v2.10.0. PaddleOCR requires specific .yml config files and pretrained models to integrate a new detection algorithm into the pipeline. Since these are missing for DB++, it is currently unavailable through out-of-the-box paddleocr.py usage.

3. Incompatibility with the current inference APIs:
   Even if DB++ is implemented somewhere in the codebase or paper, paddleocr.py uses a hardcoded list of supported detection models for simplified usage, and additional engineering work is needed to register new models properly into the pipeline, including preprocessing, postprocessing, and model loading logic.

If you want to use DB++, you would likely need to:

• Train or convert the DB++ model manually using tools/train.py or tools/infer/predict_det.py with a custom config.
• Modify the source code in paddleocr.py or related files to register DB++ manually.
• Alternatively, wait for a future release where DB++ might be supported officially, once configuration files and model weights are published.

As of now, for v2.10.0, only the officially documented detection models (such as DB, EAST, SAST, etc.) are supported in paddleocr.py.

No related issues have been identified specifically for DB++ in v2.10.0. Nonetheless, you may monitor or raise a request in GitHub Issues: https://github.com/PaddlePaddle/PaddleOCR/issues.

---

Response generated by 🤖 feifei-bot | chatgpt-4o-latest

⚠️ This is an AI-generated response. Please verify any critical information.