v0.4.3

What's Changed

• Enable package to load model from local path by @TanjaBayer in #53

New Contributors

• @TanjaBayer made their first contribution in #53

Full Changelog: 0.4.2...0.4.3

v0.4.2

What's Changed

• fix opencv version by @fcakyon in #48

Full Changelog: 0.4.1...0.4.2

v0.4.1

What's Changed

• fix crop export by @fcakyon in #45

Full Changelog: 0.4.0...0.4.1

v0.4.0

enhancement

• fix boxes outside image boundaries (#37)

breaking changes

• drop conda support, update python version (#38)

v0.3.5

• Rebuild conda binaries.

v0.3.4

• add support for PIL and numpy images in addition to filepath. #28

from PIL import Image
import numpy

# can be filepath, PIL image or numpy array
image = 'figures/idcard.png' 
image = Image.open("figures/idcard.png")
image = numpy.array(Image.open("figures/idcard.png"))

# apply craft text detection
prediction_result = craft.detect_text(image)

v0.3.3

• Relax requirements for OpenCV (#25)

v0.3.2

path for the weight file can be specified by:

load_craftnet_model(weight_path="path/to/weight")

load_refinenet_model(weight_path="path/to/weight")

release 0.3.1

• fix empty_cuda_cache

release 0.3.0

• updated basic usage for better device handling, now Craft instance should be created before calling detect_text:

# import Craft class
from craft_text_detector import Craft

# set image path and export folder directory
image_path = 'figures/idcard.png'
output_dir = 'outputs/'

# create a craft instance
craft = Craft(output_dir=output_dir, crop_type="poly", cuda=False)

# apply craft text detection and export detected regions to output directory
prediction_result = craft.detect_text(image_path)

# unload models from ram/gpu
craft.unload_craftnet_model()
craft.unload_refinenet_model()

• some internal naming and styling changes