document image degradation

Overview

ocrodeg

The ocrodeg package is a small Python library implementing document image degradation for data augmentation for handwriting recognition and OCR applications.

The following illustrates the kinds of degradations available from ocrodeg.

%pylab inline
Populating the interactive namespace from numpy and matplotlib
rc("image", cmap="gray", interpolation="bicubic")
figsize(10, 10)
import scipy.ndimage as ndi
import ocrodeg

image = imread("testdata/W1P0.png")
imshow(image)
<matplotlib.image.AxesImage at 0x7fabcc7ab390>

png

PAGE ROTATION

This is just for illustration; for large page rotations, you can just use ndimage.

for i, angle in enumerate([0, 90, 180, 270]):
    subplot(2, 2, i+1)
    imshow(ndi.rotate(image, angle))

png

RANDOM GEOMETRIC TRANSFORMATIONS

random_transform generates random transformation parameters that work reasonably well for document image degradation. You can override the ranges used by each of these parameters by keyword arguments.

ocrodeg.random_transform()
{'angle': -0.016783842893063807,
 'aniso': 0.805280370671964,
 'scale': 0.9709145529604223,
 'translation': (0.014319657859164045, 0.03676897986267606)}

Here are four samples generated by random transforms.

for i in xrange(4):
    subplot(2, 2, i+1)
    imshow(ocrodeg.transform_image(image, **ocrodeg.random_transform()))

png

You can use transform_image directly with the different parameters to get a feel for the ranges and effects of these parameters.

for i, angle in enumerate([-2, -1, 0, 1]):
    subplot(2, 2, i+1)
    imshow(ocrodeg.transform_image(image, angle=angle*pi/180))

png

for i, angle in enumerate([-2, -1, 0, 1]):
    subplot(2, 2, i+1)
    imshow(ocrodeg.transform_image(image, angle=angle*pi/180)[1000:1500, 750:1250])

png

for i, aniso in enumerate([0.5, 1.0, 1.5, 2.0]):
    subplot(2, 2, i+1)
    imshow(ocrodeg.transform_image(image, aniso=aniso))

png

for i, aniso in enumerate([0.5, 1.0, 1.5, 2.0]):
    subplot(2, 2, i+1)
    imshow(ocrodeg.transform_image(image, aniso=aniso)[1000:1500, 750:1250])

png

for i, scale in enumerate([0.5, 0.9, 1.0, 2.0]):
    subplot(2, 2, i+1)
    imshow(ocrodeg.transform_image(image, scale=scale))

png

for i, scale in enumerate([0.5, 0.9, 1.0, 2.0]):
    subplot(2, 2, i+1)
    h, w = image.shape
    imshow(ocrodeg.transform_image(image, scale=scale)[h//2-200:h//2+200, w//3-200:w//3+200])

png

RANDOM DISTORTIONS

Pages often also have a small degree of warping. This can be modeled by random distortions. Very small and noisy random distortions also model ink spread, while large 1D random distortions model paper curl.

for i, sigma in enumerate([1.0, 2.0, 5.0, 20.0]):
    subplot(2, 2, i+1)
    noise = ocrodeg.bounded_gaussian_noise(image.shape, sigma, 5.0)
    distorted = ocrodeg.distort_with_noise(image, noise)
    h, w = image.shape
    imshow(distorted[h//2-200:h//2+200, w//3-200:w//3+200])

png

RULED SURFACE DISTORTIONS

for i, mag in enumerate([5.0, 20.0, 100.0, 200.0]):
    subplot(2, 2, i+1)
    noise = ocrodeg.noise_distort1d(image.shape, magnitude=mag)
    distorted = ocrodeg.distort_with_noise(image, noise)
    h, w = image.shape
    imshow(distorted[:1500])

png

BLUR, THRESHOLDING, NOISE

There are a range of utilities for modeling imaging artifacts: blurring, noise, inkspread.

patch = image[1900:2156, 1000:1256]
imshow(patch)
<matplotlib.image.AxesImage at 0x7fabc88c7e10>

png

for i, s in enumerate([0, 1, 2, 4]):
    subplot(2, 2, i+1)
    blurred = ndi.gaussian_filter(patch, s)
    imshow(blurred)

png

for i, s in enumerate([0, 1, 2, 4]):
    subplot(2, 2, i+1)
    blurred = ndi.gaussian_filter(patch, s)
    thresholded = 1.0*(blurred>0.5)
    imshow(thresholded)

png

reload(ocrodeg)
for i, s in enumerate([0.0, 1.0, 2.0, 4.0]):
    subplot(2, 2, i+1)
    blurred = ocrodeg.binary_blur(patch, s)
    imshow(blurred)

png

for i, s in enumerate([0.0, 0.1, 0.2, 0.3]):
    subplot(2, 2, i+1)
    blurred = ocrodeg.binary_blur(patch, 2.0, noise=s)
    imshow(blurred)

png

MULTISCALE NOISE

reload(ocrodeg)
for i in range(4):
    noisy = ocrodeg.make_multiscale_noise_uniform((512, 512))
    subplot(2, 2, i+1); imshow(noisy, vmin=0, vmax=1)

png

RANDOM BLOBS

for i, s in enumerate([2, 5, 10, 20]):
    subplot(2, 2, i+1)
    imshow(ocrodeg.random_blobs(patch.shape, 3e-4, s))

png

reload(ocrodeg)
blotched = ocrodeg.random_blotches(patch, 3e-4, 1e-4)
#blotched = minimum(maximum(patch, ocrodeg.random_blobs(patch.shape, 30, 10)), 1-ocrodeg.random_blobs(patch.shape, 15, 8))
subplot(121); imshow(patch); subplot(122); imshow(blotched)
<matplotlib.image.AxesImage at 0x7fabc8a35490>

png

FIBROUS NOISE

imshow(ocrodeg.make_fibrous_image((256, 256), 700, 300, 0.01))
<matplotlib.image.AxesImage at 0x7fabc8852450>

png

FOREGROUND / BACKGROUND SELECTION

subplot(121); imshow(patch); subplot(122); imshow(ocrodeg.printlike_multiscale(patch))
<matplotlib.image.AxesImage at 0x7fabc8676d90>

png

subplot(121); imshow(patch); subplot(122); imshow(ocrodeg.printlike_fibrous(patch))
<matplotlib.image.AxesImage at 0x7fabc8d1b250>

png

Owner
NVIDIA Research Projects
NVIDIA Research Projects
The official code for the ICCV-2021 paper "Speech Drives Templates: Co-Speech Gesture Synthesis with Learned Templates".

SpeechDrivesTemplates The official repo for the ICCV-2021 paper "Speech Drives Templates: Co-Speech Gesture Synthesis with Learned Templates". [arxiv

Qian Shenhan 53 Dec 23, 2022
Image augmentation library in Python for machine learning.

Augmentor is an image augmentation library in Python for machine learning. It aims to be a standalone library that is platform and framework independe

Marcus D. Bloice 4.8k Jan 04, 2023
Ready-to-use OCR with 80+ supported languages and all popular writing scripts including Latin, Chinese, Arabic, Devanagari, Cyrillic and etc.

EasyOCR Ready-to-use OCR with 80+ languages supported including Chinese, Japanese, Korean and Thai. What's new 1 February 2021 - Version 1.2.3 Add set

Jaided AI 16.7k Jan 03, 2023
Tensorflow-based CNN+LSTM trained with CTC-loss for OCR

Overview This collection demonstrates how to construct and train a deep, bidirectional stacked LSTM using CNN features as input with CTC loss to perfo

Jerod Weinman 489 Dec 21, 2022
Text recognition (optical character recognition) with deep learning methods.

What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis | paper | training and evaluation data | failure cases and cle

Clova AI Research 3.2k Jan 04, 2023
Automatically download multiple papers by keywords in CVPR

CVFPaperHelper Automatically download multiple papers by keywords in CVPR Install mkdir PapersToRead cd PaperToRead pip install requests tqdm git clon

46 Jun 08, 2022
Omdena-abuja-anpd - Automatic Number Plate Detection for the security of lives and properties using Computer Vision.

Omdena-abuja-anpd - Automatic Number Plate Detection for the security of lives and properties using Computer Vision.

Abdulazeez Jimoh 1 Jan 01, 2022
Perspective recovery of text using transformed ellipses

unproject_text Perspective recovery of text using transformed ellipses. See full writeup at https://mzucker.github.io/2016/10/11/unprojecting-text-wit

Matt Zucker 111 Nov 13, 2022
Official PyTorch implementation for "Mixed supervision for surface-defect detection: from weakly to fully supervised learning"

Mixed supervision for surface-defect detection: from weakly to fully supervised learning [Computers in Industry 2021] Official PyTorch implementation

ViCoS Lab 169 Dec 30, 2022
Controlling the computer volume with your hands // OpenCV

HandsControll-AI Controlling the computer volume with your hands // OpenCV Step 1 git clone https://github.com/Hayk-21/HandsControll-AI.git pip instal

Hayk 1 Nov 04, 2021
The code for “Oriented RepPoints for Aerail Object Detection”

Oriented RepPoints for Aerial Object Detection The code for the implementation of “Oriented RepPoints”, Under review. (arXiv preprint) Introduction Or

WentongLi 207 Dec 24, 2022
A bot that plays TFT using OCR. Keeps track of bench, board, items, and plays the user defined team comp.

NOTES: To ensure best results, make sure you are running this on a computer that has decent specs. 1920x1080 fullscreen is required in League, game mu

francis 125 Dec 30, 2022
👄 The most accurate natural language detection library for Java and the JVM, suitable for long and short text alike

Quick Info this library tries to solve language detection of very short words and phrases, even shorter than tweets makes use of both statistical and

Peter M. Stahl 532 Dec 28, 2022
M-LSDを用いて四角形を検出し、射影変換を行うサンプルプログラム

M-LSD-warpPerspective-Example M-LSDを用いて四角形を検出し、射影変換を行うサンプルプログラムです。 Requirements OpenCV 3.4.2 or Later tensorflow 2.4.1 or Later Usage 実行方法は以下です。 pytho

KazuhitoTakahashi 9 Oct 14, 2022
Generates a message from the infamous Jerma Impostor image

Generate your very own jerma sus imposter message. Modes: Default Mode: Only supports the characters " ", !, a, b, c, d, e, h, i, m, n, o, p, q, r, s,

Giorno420 1 Oct 27, 2022
Text layer for bio-image annotation.

napari-text-layer Napari text layer for bio-image annotation. Installation You can install using pip: pip install napari-text-layer Keybindings and m

6 Sep 29, 2022
Table recognition inside douments using neural networks

TableTrainNet A simple project for training and testing table recognition in documents. This project was developed to make a neural network which reco

Giovanni Cavallin 93 Jul 24, 2022
BoxToolBox is a simple python application built around the openCV library

BoxToolBox is a simple python application built around the openCV library. It is not a full featured application to guide you through the w

František Horínek 1 Nov 12, 2021
A simple document layout analysis using Python-OpenCV

Run the application: python main.py *Note: For first time running the application, create a folder named "output". The application is a simple documen

Roinand Aguila 109 Dec 12, 2022
Virtualdragdrop - Virtual Drag and Drop Using OpenCV and Arduino

Virtualdragdrop - Virtual Drag and Drop Using OpenCV and Arduino

Rizky Dermawan 4 Mar 10, 2022