ID Card Detector

Everything you need for ID card detection. Easily perform ID card detection/segmentation, fit quad to detected mask and unwarp the ID card surface to rectangle with Pytorch models on Linux&Windows. Works on all types of ID cards.

Package maintainer: Fatih Cagatay Akyon

Overview

Perform instance segmentation for ID cards with pretrained MaskRCNN model and get predicted bounding boxes, masks and prediction scores. Works on CPU/GPU and Linux/Windows.

Getting started

Installation

pip install id-card-detector

Basic Usage

# import package
import id_card_detector

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

# perform id card detection and export results to output directory
_,_,_,_ = id_card_detector.detect_card(image_path=image_path,
				       output_dir=output_dir
				       unwarp=True)

Advanced Usage

# import package
import id_card_detector as card_det

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

# read image
image = card_det.read_image(image_path)

# get prediction
masks, boxes, classes, scores = card_det.get_prediction(image=image, threshold=0.75)

# export a visual of detected bboxes and masks
prediction_visual = card_det.visualize_prediction(image, masks, boxes, classes,
                                                  rect_th=2,
                                           	  text_size=0.85,
                                         	  text_th=2,
                                         	  color=color,
                                         	  output_dir=output_dir)

# export detected bounding boxes
card_det.export_predicted_bboxes(image=image,
                        	 boxes=boxes,
                        	 output_dir=output_dir)

# fit quads to predicted masks
quads = card_det.fit_quads_over_masks(image, masks)

# visualize/export quads
quad_visual = card_det.visualize_quads(image=image,
				       quads=quads,
				       output_dir=output_dir,
                              	       color=color)

# unwarp quads to rects
unwarped_quads = card_det.unwarp_quads(image, quads)

# export unwarped quads
card_det.export_unwarped_quads(unwarped_quads,
                               output_dir=output_dir)