Upload 2 files

README.md

@@ -0,0 +1,102 @@
+---
+title: KYC OCR (India)
+emoji: 📄
+colorFrom: blue
+colorTo: green
+sdk: gradio
+sdk_version: "4.44.0"
+python_version: "3.12"
+app_file: app.py
+pinned: false
+license: mit
+short_description: Extract fields from Indian PAN cards and cheques on CPU
+---
+
+# 🇮🇳 KYC Document OCR
+
+A lightweight, CPU-friendly pipeline for extracting structured fields from
+Indian KYC documents (PAN card, cheque) from images or PDFs.
+
+Built on [RapidOCR](https://github.com/RapidAI/RapidOCR) — ONNX-runtime
+deployment of PaddleOCR PP-OCRv4 models. No GPU needed.
+
+## What it does
+
+- **PDF or image input** (PyMuPDF rasterises PDFs at 300 DPI)
+- **Document type auto-detection** via keyword + regex scoring
+- **Cheque rotation correction** — detects sideways-scanned cheques (text
+  spans more vertically than horizontally) and rotates before MICR extraction
+- **Format-spec validation** — PAN regex with 4th-character entity-type
+  lookup, IFSC regex, date normalisation, MICR positional parse
+
+## Supported documents
+
+| Document | Fields extracted |
+|---|---|
+| **PAN card** | PAN number, name, father's name (individual) / company name, DOB / date of incorporation, entity type (Individual / Company / HUF / Firm / Trust / etc.) |
+| **Cheque** | Bank name, branch address, IFSC code, account number, 9-digit MICR code, cheque number, transaction code |
+| **Aadhaar** | _Coming soon — recommended path: read the secure QR code via `pyzbar` rather than OCR_ |
+| **Passport** | _Coming soon — recommended path: parse the MRZ via `passporteye`_ |
+
+## How to use the Space
+
+1. Drop a PDF or image into the upload box
+2. Click **Extract fields**
+3. View the structured output and raw OCR (for debugging)
+
+The included example is a synthetic PAN with fake data so you can try it
+without uploading real documents.
+
+## Running locally
+
+```bash
+git clone <this-space-url>
+cd <repo>
+pip install -r requirements.txt
+python app.py
+```
+
+Then open `http://127.0.0.1:7860`.
+
+## Using `kyc_extract.py` directly as a library
+
+```python
+from kyc_extract import process, RapidOCR
+
+engine = RapidOCR()
+results = process("path/to/document.pdf", engine)
+
+for page in results:
+    print(page["detected_type"])     # 'pan' or 'cheque'
+    print(page["fields"])             # structured dict
+```
+
+## Privacy
+
+This is a public Space. **Do not upload real personal documents.** For
+production KYC pipelines, fork this repo and deploy on your own private
+infrastructure (DPDP Act and UIDAI guidelines require it for Aadhaar in
+particular).
+
+## Architecture notes
+
+- All extraction is rule-based — no model is hardcoded to specific names,
+  banks, PANs, or addresses. The same script handles any PAN or any bank's
+  cheque.
+- Document classification uses keyword + regex scoring; if confidence is
+  low, type is `unknown` and no extractor runs.
+- For cheques in portrait-orientation PDFs (common when phones are held
+  vertically), the pipeline detects the layout anomaly and re-OCRs the
+  rotated image, merging text from both passes so address and IFSC don't
+  get lost in the rotation.
+
+## Limitations
+
+- English-only OCR — Hindi label text (`नाम/Name`) comes out garbled but
+  the pipeline has shape-based fallbacks so this doesn't break field
+  extraction. For full Hindi extraction, add `lang='hi'` to RapidOCR.
+- MICR fonts are E-13B and not perfectly read by general OCR. The pipeline
+  reconstructs the 9-digit MICR from OCR fragments via positional parsing,
+  which works on most cheques but degrades on poor scans.
+- Aadhaar and passport extractors are stubs. The proper approach uses QR
+  and MRZ respectively, not pure OCR.

app.py

@@ -0,0 +1,149 @@
+"""
+KYC OCR — Gradio app for Hugging Face Spaces.
+
+Upload a PAN card, cheque, Aadhaar, or passport (PDF or image) and get
+structured field extraction. Runs on CPU using RapidOCR (ONNX).
+"""
+import gradio as gr
+import time
+import traceback
+from pathlib import Path
+
+from kyc_extract import process, RapidOCR
+
+# ── Initialise OCR engine once at startup ─────────────────────────────
+print("Loading RapidOCR engine (ONNX models)…")
+_t0 = time.time()
+ENGINE = RapidOCR()
+print(f"Ready. Engine init took {time.time() - _t0:.1f}s")
+
+
+# ── Pretty-format the extracted fields as Markdown ────────────────────
+def _fmt_fields(fields: dict) -> str:
+    if not fields:
+        return "_No fields extracted._"
+    lines = []
+    for k, v in fields.items():
+        label = k.replace("_", " ").title()
+        if isinstance(v, bool):
+            v = "✅ Yes" if v else "❌ No"
+        elif v is None:
+            v = "_(not found)_"
+        lines.append(f"**{label}:** {v}")
+    return "\n\n".join(lines)
+
+
+def _fmt_info(result: dict) -> str:
+    bits = [
+        f"**Detected type:** `{result.get('detected_type', 'unknown').upper()}`",
+        f"**OCR time:** {result.get('ocr_time_s')}s",
+        f"**Rotation applied:** {result.get('rotation_applied', 0)}°",
+        f"**OCR quality score:** {result.get('ocr_score')}",
+    ]
+    return "  •  ".join(bits)
+
+
+# ── Main handler ──────────────────────────────────────────────────────
+def extract(file):
+    if file is None:
+        return "_Please upload a file._", "", ""
+
+    file_path = file if isinstance(file, str) else file.name
+
+    try:
+        pages = process(file_path, ENGINE)
+    except Exception as exc:
+        return (
+            f"❌ **Error:** {exc}\n\n```\n{traceback.format_exc()}\n```",
+            "",
+            "",
+        )
+
+    if not pages:
+        return "_No pages processed._", "", ""
+
+    # Render all pages (cheques and PANs are normally single-page; multi-page
+    # PDFs of compound KYC packets get each page rendered separately).
+    info_blocks, field_blocks, raw_blocks = [], [], []
+    for p in pages:
+        if "error" in p:
+            info_blocks.append(f"### Page {p['page']}\n_{p['error']}_")
+            continue
+        info_blocks.append(f"### Page {p['page']}\n{_fmt_info(p)}")
+        field_blocks.append(f"### Page {p['page']}\n{_fmt_fields(p['fields'])}")
+        raw_blocks.append(
+            f"━━━ Page {p['page']} ━━━\n"
+            + "\n".join(f"[{c:.2f}] {t}" for c, t in p["_raw_lines"])
+        )
+
+    return (
+        "\n\n".join(info_blocks),
+        "\n\n---\n\n".join(field_blocks) if field_blocks else "_No fields extracted._",
+        "\n\n".join(raw_blocks),
+    )
+
+
+# ── Gradio UI ─────────────────────────────────────────────────────────
+DESCRIPTION = """
+# 🇮🇳 KYC Document OCR
+
+Extract structured fields from Indian KYC documents — **PAN card**, **cheque**,
+**Aadhaar**, and **passport** — from images or PDFs. Runs entirely on CPU using
+[RapidOCR](https://github.com/RapidAI/RapidOCR) (PaddleOCR PP-OCRv4 models in ONNX).
+
+**Supported documents and fields:**
+- **PAN:** PAN number, name, father's name, DOB / date of incorporation, entity type
+- **Cheque:** bank name, address, IFSC, account number, MICR code, cheque number, transaction code
+- **Aadhaar / Passport:** _coming soon — Aadhaar reads best from QR code (`pyzbar`), passport from MRZ (`passporteye`)_
+
+**Privacy note:** this is a public Space. Don't upload real personal documents.
+For production use, fork it and run on private infrastructure.
+"""
+
+with gr.Blocks(title="KYC OCR") as demo:
+    gr.Markdown(DESCRIPTION)
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            file_input = gr.File(
+                label="Upload PAN / Cheque (PDF or image)",
+                file_types=[".pdf", ".png", ".jpg", ".jpeg", ".webp"],
+                type="filepath",
+            )
+            submit_btn = gr.Button("Extract fields", variant="primary", size="lg")
+            gr.Examples(
+                examples=[["examples/synthetic_pan.png"]],
+                inputs=file_input,
+                label="Example (synthetic PAN with fake data)",
+            )
+
+        with gr.Column(scale=2):
+            info_out = gr.Markdown()
+            fields_out = gr.Markdown()
+
+    with gr.Accordion("Raw OCR output (for debugging)", open=False):
+        raw_out = gr.Textbox(label="Raw OCR lines with confidence scores",
+                             lines=20)
+
+    gr.Markdown(
+        "---\n"
+        "Pipeline source: `kyc_extract.py`. "
+        "All extraction is rule-based — no hardcoded names/banks. "
+        "Doc-type detection → optional rotation correction (cheques) → "
+        "format-spec regex + heuristics + checksum validation."
+    )
+
+    submit_btn.click(
+        extract,
+        inputs=file_input,
+        outputs=[info_out, fields_out, raw_out],
+    )
+    file_input.change(
+        extract,
+        inputs=file_input,
+        outputs=[info_out, fields_out, raw_out],
+    )
+
+
+if __name__ == "__main__":
+    demo.launch(theme=gr.themes.Soft())