## What to put here

This folder is served by `web/index.html`.

### Required (recognition demo)

- `ocr_rec.onnx`: exported recognition model.
- `ppocr_keys_v1.txt`: vocabulary file (UTF-8, one character per line) for greedy CTC decoding.

The web demo expects:

- `./models/ocr_rec.onnx`
- `./models/ppocr_keys_v1.txt`

### Optional (for full OCR later)

- `ocr_det.onnx`: exported detection model (not yet used by the current minimal demo).

## How to export ONNX here

Put PaddleOCR v6 tiny inference folders under:

`models/paddleocr_v6_tiny/`

Expected layout:

- `models/paddleocr_v6_tiny/rec/`
  - `inference.json` (or `inference.pdmodel` / `model.json`)
  - `inference.pdiparams`
- `models/paddleocr_v6_tiny/det/`
  - `inference.json` (or `inference.pdmodel` / `model.json`)
  - `inference.pdiparams`

Then run the VS Code debug config:

`Export PaddleOCR v6 tiny (det+rec) → ONNX`

which writes `web/models/ocr_rec.onnx` and `web/models/ocr_det.onnx`.