feat: Migrate VIN extractor to engine abstraction (#115) #117

New Issue

2026-02-07T16:12:52Z

egullickson commented

2026-02-07 16:12:52 +00:00

Relates to #115

Migrate vin_extractor.py and ocr_service.py to use the new OcrEngine abstraction instead of calling pytesseract directly. PaddleOCR becomes primary engine.

Changes

Modify ocr/app/extractors/vin_extractor.py - Replace pytesseract.image_to_data() calls with engine.recognize()
Modify ocr/app/services/ocr_service.py - Replace pytesseract calls with engine interface
Update ocr/app/extractors/receipt_extractor.py - Same migration (if uses Tesseract directly)
Preserve VIN preprocessing pipeline (vin_preprocessor.py unchanged)
Preserve VIN validation pipeline (vin_validator.py unchanged)

Acceptance Criteria

VIN extractor uses engine.recognize() instead of pytesseract directly
Generic OCR service uses engine interface
PSM mode fallback strategy adapted for PaddleOCR (angle detection replaces PSM modes)
VIN character whitelist equivalent implemented for PaddleOCR
Confidence scoring works with PaddleOCR output format
Receipt and manual extraction endpoints still function (no regression)

Relates to #115 Migrate `vin_extractor.py` and `ocr_service.py` to use the new OcrEngine abstraction instead of calling pytesseract directly. PaddleOCR becomes primary engine. ## Changes - Modify `ocr/app/extractors/vin_extractor.py` - Replace `pytesseract.image_to_data()` calls with `engine.recognize()` - Modify `ocr/app/services/ocr_service.py` - Replace `pytesseract` calls with engine interface - Update `ocr/app/extractors/receipt_extractor.py` - Same migration (if uses Tesseract directly) - Preserve VIN preprocessing pipeline (vin_preprocessor.py unchanged) - Preserve VIN validation pipeline (vin_validator.py unchanged) ## Acceptance Criteria - [ ] VIN extractor uses engine.recognize() instead of pytesseract directly - [ ] Generic OCR service uses engine interface - [ ] PSM mode fallback strategy adapted for PaddleOCR (angle detection replaces PSM modes) - [ ] VIN character whitelist equivalent implemented for PaddleOCR - [ ] Confidence scoring works with PaddleOCR output format - [ ] Receipt and manual extraction endpoints still function (no regression)

egullickson added the

labels 2026-02-07 16:13:23 +00:00

egullickson added this to the Sprint 2026-02-02 milestone 2026-02-07 16:13:31 +00:00

egullickson referenced this issue

2026-02-07 16:14:33 +00:00

feat: Improve OCR process - replace Tesseract with PaddleOCR and add cloud fallback for VIN scanning #115

egullickson referenced this issue

2026-02-07 16:22:19 +00:00