himipo/ocr-workbench-zerogpu-ppocrv6

PP-OCRv6 Medium Worker

This worker deliberately combines two models:

• —PP-OCRv6_medium_det for text detection
• —PP-OCRv6_medium_rec for text recognition

The user-specified PP-OCRv6_medium_det_safetensors repository is detection-only, so it cannot produce the requested text result by itself.

The default Dockerfile is CPU-oriented because the complete PP-OCRv6 medium pipeline is lightweight. For GPU deployment, replace Dockerfile with Dockerfile.gpu before uploading the Space.

Variables

• —TEXT_DETECTION_MODEL=PP-OCRv6_medium_det
• —TEXT_RECOGNITION_MODEL=PP-OCRv6_medium_rec
• —PPOCR_ENGINE=transformers
• —PADDLEOCR_DEVICE=cpu (or gpu:0 with the GPU Dockerfile)
• —WORKER_API_TOKEN as an optional shared secret.