alessandrozamberletti / docdetect

Real-time detection of documents in images

docdetect

Simple real-time detection of documents in images using Canny Edge Detection, Hough Transform and Depth First Search.

Installation

To install, use pip or easy_install:

$ pip install --upgrade docdetect

or

$ easy_install --upgrade docdetect

Instructions

TBD

Examples

Process an image:

import docdetect

rects = docdetect.process(image)
image = docdetect.draw(rects, image)

Process a video:

import cv2
import docdetect

video = cv2.VideoCapture(video_path)
cv2.startWindowThread()
cv2.namedWindow('output')
while video.isOpened():
    ret, frame = video.read()
    if ret:
        rects = docdetect.process(frame)
        frame = docdetect.draw(rects, frame)
        cv2.imshow('output', frame)
        cv2.waitKey(1)
video.release()

Resources

• Fast and Accurate Document Detection for Scanning

docdetect's Issues

PyPI Package Syntax Error

The line mser = cv2.MSER_create(_max_area=max_area) under the _find_characters(im, max_area): function in canny_edges.py is incorrect. It should be mser = cv2.MSER_create(max_area=max_area) or the code fails to work.

Could you tell me about the labeled dataset , in Document Detection Article?

HI,It is a good and complex job.cool.

I tried many methods to overcome complex photos.
But ,I can not find a labeled dataset .
I made a small dataset, but it dose not work well in new photos.

Could you give the best size of input image?

When I run the detect program,it's ok for size[500,500],but can't get result for size[224,224](use the same image)