Train your own model

If the pretrained models don’t meet your specific needs, you have the option to train your own model using the docTR library. For details on the training process and the necessary data and data format, refer to the following links:

• detection

• recognition

Loading your custom trained model

This section shows how you can easily load a custom trained model in docTR.

TensorFlowPyTorch

from doctr.models import ocr_predictor, db_resnet50, crnn_vgg16_bn

# Load custom detection model
det_model = db_resnet50(pretrained=False, pretrained_backbone=False)
det_model.load_weights("<path_to_checkpoint>")
predictor = ocr_predictor(det_arch=det_model, reco_arch="vitstr_small", pretrained=True)

# Load custom recognition model
reco_model = crnn_vgg16_bn(pretrained=False, pretrained_backbone=False)
reco_model.load_weights("<path_to_checkpoint>")
predictor = ocr_predictor(det_arch="linknet_resnet18", reco_arch=reco_model, pretrained=True)

# Load custom detection and recognition model
det_model = db_resnet50(pretrained=False, pretrained_backbone=False)
det_model.load_weights("<path_to_checkpoint>")
reco_model = crnn_vgg16_bn(pretrained=False, pretrained_backbone=False)
reco_model.load_weights("<path_to_checkpoint>")
predictor = ocr_predictor(det_arch=det_model, reco_arch=reco_model, pretrained=False)

Load a custom recognition model trained on another vocabulary as the default one (French):

TensorFlowPyTorch

from doctr.models import ocr_predictor, crnn_vgg16_bn
from doctr.datasets import VOCABS

reco_model = crnn_vgg16_bn(pretrained=False, pretrained_backbone=False, vocab=VOCABS["german"])
reco_model.load_weights("<path_to_checkpoint>")

predictor = ocr_predictor(det_arch='linknet_resnet18', reco_arch=reco_model, pretrained=True)

Load a custom trained KIE detection model:

TensorFlowPyTorch

from doctr.models import kie_predictor, db_resnet50

det_model = db_resnet50(pretrained=False, pretrained_backbone=False, class_names=['total', 'date'])
det_model.load_weights("<path_to_checkpoint>")
kie_predictor(det_arch=det_model, reco_arch='crnn_vgg16_bn', pretrained=True)

Load a model with customized Preprocessor:

TensorFlowPyTorch

from doctr.models.predictor import OCRPredictor
from doctr.models.detection.predictor import DetectionPredictor
from doctr.models.recognition.predictor import RecognitionPredictor
from doctr.models.preprocessor import PreProcessor
from doctr.models import db_resnet50, crnn_vgg16_bn

det_model = db_resnet50(pretrained=False, pretrained_backbone=False)
det_model.load_weights("<path_to_checkpoint>")
reco_model = crnn_vgg16_bn(pretrained=False, pretrained_backbone=False)
reco_model.load_weights("<path_to_checkpoint>")

det_predictor = DetectionPredictor(
    PreProcessor(
        (1024, 1024),
        batch_size=1,
        mean=(0.798, 0.785, 0.772),
        std=(0.264, 0.2749, 0.287)
    ),
    det_model
)

reco_predictor = RecognitionPredictor(
    PreProcessor(
        (32, 128),
        preserve_aspect_ratio=True,
        batch_size=32,
        mean=(0.694, 0.695, 0.693),
        std=(0.299, 0.296, 0.301)
    ),
    reco_model
)

predictor = OCRPredictor(det_predictor, reco_predictor)

Custom orientation classification models

If you work with rotated documents and make use of the orientation classification feature by passing one of the following arguments:

• assume_straight_pages=False

• detect_orientation=True

• straigten_pages=True

You can train your own orientation classification model using the docTR library. For details on the training process and the necessary data and data format, refer to the following link:

• orientation

NOTE: Currently we support only mobilenet_v3_small models for crop and page orientation classification.

Loading your custom trained orientation classification model

TensorFlowPyTorch

from doctr.io import DocumentFile
from doctr.models import ocr_predictor, mobilenet_v3_small_page_orientation, mobilenet_v3_small_crop_orientation
from doctr.models.classification.zoo import crop_orientation_predictor, page_orientation_predictor

custom_page_orientation_model = mobilenet_v3_small_page_orientation(pretrained=False)
custom_page_orientation_model.load_weights("<path_to_checkpoint>")
custom_crop_orientation_model = mobilenet_v3_small_crop_orientation(pretrained=False)
custom_crop_orientation_model.load_weights("<path_to_checkpoint>")

predictor = ocr_predictor(
    pretrained=True,
    assume_straight_pages=False,
    straighten_pages=True,
    detect_orientation=True,
)

# Overwrite the default orientation models
predictor.crop_orientation_predictor = crop_orientation_predictor(custom_crop_orientation_model)
predictor.page_orientation_predictor = page_orientation_predictor(custom_page_orientation_model)