helpy4/immich
1
0
Fork 0
mirror of https://github.com/immich-app/immich synced 2025-11-07 17:27:20 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat: enhance OCR configuration and functionality

Browse source 
- Updated OCR settings to include minimum detection box score, minimum detection score, and minimum recognition score.
- Refactored PaddleOCRecognizer to utilize new scoring parameters.
- Introduced new database tables for asset OCR data and search functionality.
- Modified related services and repositories to support the new OCR features.
- Updated translations for improved clarity in settings UI.
This commit is contained in:
CoderKang 2025-06-03 16:19:50 +08:00 • committed by mertalev
parent df36a09cd3
commit 4d8e51ede6
No known key found for this signature in database
GPG key ID: DF6ABC77AAD98C95
17 changed files with 180 additions and 51 deletions
Download patch file Download diff file Expand all files Collapse all files

31
machine-learning/immich_ml/models/ocr/paddle.py
View file

@ -12,10 +12,9 @@ class PaddleOCRecognizer(InferenceModel):
depends = []
identity = (ModelType.OCR, ModelTask.OCR)
def __init__(self, model_name: str, min_score: float = 0.9, **model_kwargs: Any) -> None:
self.min_score = model_kwargs.pop("minScore", min_score)
self.orientation_classify_enabled = model_kwargs.pop("orientationClassifyEnabled", True)
self.unwarping_enabled = model_kwargs.pop("unwarpingEnabled", True)
def __init__(self, model_name: str, **model_kwargs: Any) -> None:
self.orientation_classify_enabled = model_kwargs.get("orientationClassifyEnabled", False)
self.unwarping_enabled = model_kwargs.get("unwarpingEnabled", False)
super().__init__(model_name, **model_kwargs)
self._load()
self.loaded = True
@ -28,23 +27,25 @@ class PaddleOCRecognizer(InferenceModel):
use_doc_unwarping=self.unwarping_enabled,
)
def configure(self, **kwargs: Any) -> None:
self.min_detection_score = kwargs.get("minDetectionScore", 0.3)
self.min_detection_box_score = kwargs.get("minDetectionBoxScore", 0.6)
self.min_recognition_score = kwargs.get("minRecognitionScore", 0.0)
def _predict(self, inputs: NDArray[np.uint8] | bytes | Image.Image, **kwargs: Any) -> List[OCROutput]:
inputs = decode_cv2(inputs)
results = self.model.predict(inputs)
valid_texts_and_scores = [
(text, score, box)
for result in results
for text, score, box in zip(result['rec_texts'], result['rec_scores'], result['rec_polys'])
if score >= self.min_score
]
if not valid_texts_and_scores:
return []
results = self.model.predict(
inputs,
text_det_thresh=self.min_detection_score,
text_det_box_thresh=self.min_detection_box_score,
text_rec_score_thresh=self.min_recognition_score
)
return [
OCROutput(
text=text, confidence=score,
x1=box[0][0], y1=box[0][1], x2=box[1][0], y2=box[1][1],
x3=box[2][0], y3=box[2][1], x4=box[3][0], y4=box[3][1]
)
for text, score, box in valid_texts_and_scores
for result in results
for text, score, box in zip(result['rec_texts'], result['rec_scores'], result['rec_polys'])
]