gocr

package module
v0.0.0-...-fde7ae2 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Mar 4, 2024 License: Apache-2.0 Imports: 20 Imported by: 0

README

gocr

gocr is a go based OCR module

Available Scanner

  • Linear Mean Scanner
  • Circular Scanner (Currently the best)

Available predictor

  • NNPredictor (kNN with k = 1)
  • CNNPredictor (Need to install tensorflow first) (Doesn't upport custom train)

Future Improvement

  • Sauvola Segmentation
  • Better character detection
  • Better symbol detection
  • Spelling Correction

How to use

To use OCR we have some generated Model that we train using tensorflow. We use Convolutional Neural Network as the model architecture and produce good detection.

You can load and use the model to CNN Predictor like this example below:

d, _ := os.Getwd()

image, _ := gocr.ReadImage(d + "/imagetext_3.png")
s := gocr.NewCNNPredictorFromDir(modelPath + "tensor_4/")

// Define the image size
s.InputHeight, s.InputWidth = 64, 64

strings := gocr.ScanToStrings(s, image)
for _, s := range strings {
  fmt.Println(s)
}

However you can also use your own train data. Currently the predictor that support custom training only NNPredictor. Training takes csv file that have file image path and string representation.

ie:

a.gif,a
b.gif,b
A1.gif,A
B1.gif,B

Train the sample data to model and scan image

d, _ := os.Getwd()

// Load the sample data and save it in model.cbor file
err := gocr.TrainAverage(d+"/English/Fnt/", d+"/English/")
if err != nil {
  panic(err)
}

image, _ := gocr.ReadImage(d + "/imagetext_3.png")
s := gocr.NewCNNPredictorFromDir(modelPath + "sample1/model.cbor")

strings := gocr.ScanToStrings(s, image)
for _, s := range strings {
  fmt.Println(s)
}

License

gocr is released under the Apache 2.0 License. se LICENSE for details.

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

func CirucularScan 

func CirucularScan(image ImageMatrix) ([][]*Square, [][]ImageMatrix)

func Col2im 

func Col2im(x *mat64.Dense, d, r, c int) []*mat64.Dense

func Convolve 

func Convolve(x, k [][]*mat64.Dense, p, s int) [][]*mat64.Dense

func EuclideanDistance 

func EuclideanDistance(m1, m2 ImageMatrix) float64

Find the distance of 2 give Dense using Euclidean Distance

func Im2col 

func Im2col(x []*mat64.Dense, p, s, kd, kr, kc int) *mat64.Dense

func ImageMatrixToImage 

func ImageMatrixToImage(imageArray ImageMatrix, outPath string, mul int) error

Convert ImageMatrix to Image and save it to given path

func LeakyRelu 

func LeakyRelu(x float64, deriv bool) float64

func LinearScan 

func LinearScan(data ImageMatrix) [][]ImageMatrix

Scan the DarkSquare and return it as []mat64.Dense for each line And [][]mat64.Dense for each characther

func MaxPool 

func MaxPool(x *mat64.Dense, fr, fc int) (*mat64.Dense, *mat64.Dense)

func Prepare 

func Prepare(targetPath string) error

Download the dataset from here and save it in chars74k_dataset/EnglishFnt Also copy index.csv to extracted folder in targetPath

func ReadCSV 

func ReadCSV(path string) ([][]string, error)

Read CSV from given path and return array [][]string

func ReadImage 

func ReadImage(path string) (image.Image, error)

Read image in given path Can open and decode some type of image (.png, .jpg, .gif)

func Relu 

func Relu(x float64, deriv bool) float64

func Reshape2DTo2D 

func Reshape2DTo2D(x *mat64.Dense, r, c int) *mat64.Dense

func ReshapeKernel 

func ReshapeKernel(x [][]*mat64.Dense) *mat64.Dense

func ScanToStrings 

func ScanToStrings(p Predictor, image image.Image) []string

func Sigmoid 

func Sigmoid(x float64, deriv bool) float64

func Train 

func Train(sampleFolderPath string, modelPath string) error

Train read the image file from sample path convert it to model and save it in given model path The train folder should include index.csv and images that inside the index.csv

Types

type CNNPredictor 

type CNNPredictor struct {
	InputHeight int
	InputWidth  int
	// contains filtered or unexported fields
}

func NewCNNPredictor 

func NewCNNPredictor(graph *tf.Graph, labels []string) *CNNPredictor

func NewCNNPredictorFromDir 

func NewCNNPredictorFromDir(dir string) *CNNPredictor

func (*CNNPredictor) Predicts 

func (p *CNNPredictor) Predicts(images ImageMatrixs) []string

type Coordinate 

type Coordinate struct {
	// contains filtered or unexported fields
}

func NewCoordinate 

func NewCoordinate(r, c int) *Coordinate

func (*Coordinate) DistanceTo 

func (c *Coordinate) DistanceTo(c2 *Coordinate) float64

func (*Coordinate) E 

func (c *Coordinate) E() *Coordinate

func (*Coordinate) HorizontalDistanceTo 

func (c *Coordinate) HorizontalDistanceTo(c2 *Coordinate) float64

func (*Coordinate) IsInside 

func (c *Coordinate) IsInside(s *Square) bool

func (*Coordinate) N 

func (c *Coordinate) N() *Coordinate

func (*Coordinate) NE 

func (c *Coordinate) NE() *Coordinate

func (*Coordinate) NW 

func (c *Coordinate) NW() *Coordinate

func (*Coordinate) S 

func (c *Coordinate) S() *Coordinate

func (*Coordinate) SE 

func (c *Coordinate) SE() *Coordinate

func (*Coordinate) SW 

func (c *Coordinate) SW() *Coordinate

func (*Coordinate) VerticalDistanceTo 

func (c *Coordinate) VerticalDistanceTo(c2 *Coordinate) float64

func (*Coordinate) W 

func (c *Coordinate) W() *Coordinate

type Groups 

type Groups struct {
	// contains filtered or unexported fields
}

type ImageMatrix 

type ImageMatrix [][]uint8

func AdaptiveThres 

func AdaptiveThres(im ImageMatrix, bs int) ImageMatrix

func ImageToGraysclaeArray 

func ImageToGraysclaeArray(src image.Image) ImageMatrix

Convert image to grayscale 2D array

func NewImageMatrix 

func NewImageMatrix(r, c int) ImageMatrix

func NewImageMatrixWithDefaultValue 

func NewImageMatrixWithDefaultValue(r, c int, v uint8) ImageMatrix

func OtsuThresh 

func OtsuThresh(im ImageMatrix) ImageMatrix

Thresholding Using Otsu's Method

func PadAndResize 

func PadAndResize(matrix ImageMatrix, dr, dc int) ImageMatrix

Adding pad to make a square matrix Then resize it to given row length and column length

func SauvolaBinarization 

func SauvolaBinarization(imageArr ImageMatrix) ImageMatrix

Binarize the given imageArr using Best algorithm based on this paper https://pdfs.semanticscholar.org/6347/5461213fdaa24e418c33454c72bdbbe8f8b4.pdf is Sauvola Sauvola Reference: http://www.mediateam.oulu.fi/publications/pdf/24.p TODO: Implement Sauvola algorithm

func Threshold 

func Threshold(im ImageMatrix, thrs uint8) ImageMatrix

Basic Thresholding

func (ImageMatrix) At 

func (i ImageMatrix) At(r, c int) uint8

func (ImageMatrix) AtCoordinate 

func (i ImageMatrix) AtCoordinate(c *Coordinate) uint8

func (ImageMatrix) Col 

func (i ImageMatrix) Col(c int) ImageVector

func (ImageMatrix) Dims 

func (i ImageMatrix) Dims() (int, int)

func (ImageMatrix) Erode 

func (im ImageMatrix) Erode()

Erode using 4 Neighborhood

func (ImageMatrix) Historgram 

func (im ImageMatrix) Historgram() []int

func (ImageMatrix) NNInterpolation 

func (i ImageMatrix) NNInterpolation(tr, tc int) ImageMatrix

func (ImageMatrix) Pad 

func (i ImageMatrix) Pad(top, bottom, left, right int, value uint8) ImageMatrix

func (ImageMatrix) Row 

func (i ImageMatrix) Row(r int) ImageVector

func (ImageMatrix) Set 

func (i ImageMatrix) Set(r, c int, value uint8)

func (ImageMatrix) SetSquare 

func (im ImageMatrix) SetSquare(s *Square, si ImageMatrix)

func (ImageMatrix) Slice 

func (i ImageMatrix) Slice(sr, er, sc, ec int) ImageMatrix

func (ImageMatrix) SliceSquare 

func (im ImageMatrix) SliceSquare(s *Square) ImageMatrix

type ImageMatrixs 

type ImageMatrixs []ImageMatrix

func (ImageMatrixs) Average 

func (is ImageMatrixs) Average() ImageMatrix

type ImageVector 

type ImageVector []uint8

func NewImageVector 

func NewImageVector(l int) ImageVector

func (ImageVector) Sum 

func (v ImageVector) Sum() uint64

type Marker 

type Marker struct {
	Start int
	End   int
}

func MarkersOfMatrix 

func MarkersOfMatrix(data ImageMatrix, threshold float64, direction int) []Marker

Find Marker of DarkSquare from given Matrix Direction 0 means it will iterate every rows Direction 1 means it will iterate every columns

type Model 

type Model struct {
	Name        string
	ModelImages []ModelImage
}

type ModelImage 

type ModelImage struct {
	Label string
	Data  ImageMatrix
}

type NNPredictor 

type NNPredictor struct {
	// contains filtered or unexported fields
}

================================= Nearest Neighbor Predictor =================================

func NewNNPredictor 

func NewNNPredictor(model *Model) *NNPredictor

func NewNNPredictorFromFile 

func NewNNPredictorFromFile(path string) *NNPredictor

func (*NNPredictor) Predicts 

func (p *NNPredictor) Predicts(images ImageMatrixs) []string

type Predictor 

type Predictor interface {
	Predicts(ImageMatrixs) []string
	// contains filtered or unexported methods
}

type Square 

type Square struct {
	// contains filtered or unexported fields
}

func NewSquare 

func NewSquare(coor1, coor2 *Coordinate) *Square

func (*Square) Area 

func (s *Square) Area() int

func (*Square) AverageDistanceTo 

func (s *Square) AverageDistanceTo(s2 *Square) float64

func (*Square) AverageHorizontalDistanceTo 

func (s *Square) AverageHorizontalDistanceTo(s2 *Square) float64

func (*Square) AverageVerticalDistanceTo 

func (s *Square) AverageVerticalDistanceTo(s2 *Square) float64

func (*Square) DistancesTo 

func (s *Square) DistancesTo(s2 *Square) []float64

func (*Square) Expand 

func (s *Square) Expand(c *Coordinate)

func (*Square) FarthestDistanceTo 

func (s *Square) FarthestDistanceTo(s2 *Square) float64

func (*Square) FarthestHorizontalDistanceTo 

func (s *Square) FarthestHorizontalDistanceTo(s2 *Square) float64

func (*Square) FarthestVerticalDistanceTo 

func (s *Square) FarthestVerticalDistanceTo(s2 *Square) float64

func (*Square) Height 

func (s *Square) Height() int

func (*Square) HorizontalDistancesTo 

func (s *Square) HorizontalDistancesTo(s2 *Square) []float64

func (*Square) Include 

func (s *Square) Include(r, c int) bool

func (*Square) Merge 

func (s *Square) Merge(a2 *Square)

func (*Square) NearestDistanceTo 

func (s *Square) NearestDistanceTo(s2 *Square) float64

func (*Square) NearestHorizontalDistanceTo 

func (s *Square) NearestHorizontalDistanceTo(s2 *Square) float64

func (*Square) NearestVerticalDistanceTo 

func (s *Square) NearestVerticalDistanceTo(s2 *Square) float64

func (*Square) VerticalDistancesTo 

func (s *Square) VerticalDistancesTo(s2 *Square) []float64

func (*Square) Width 

func (s *Square) Width() int

Source Files

View all Source files

Directories

Path Synopsis
examples
basic
convolution

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.