can it generate this type image? about textrecognitiondatagenerator

[Belval]

At the moment no, but I will try to implement it this weekend.

[aagainbeyond2]

what's more , could it generate bold or italic text like WPS do?

[Belval]

Bold/italic/normal is managed as the font level so you'd want the italic/bold version of whichever font you are using.

Also, what is WPS?

[hqzz]

WPS is a chinese office software like MS word@Belval

[Belval]

Thanks!

[hongyinjie]

italic can done by next step:

def italic_img(img, angle):
    ih, iw = img.shape[:2]
    dw = int(math.ceil(ih * math.tan(math.pi * angle / 360)))
    if dw > 0:
        img1 = cv2.copyMakeBorder(img, 0 , 0, dw, dw, cv2.BORDER_CONSTANT, value=[255, 255, 255])
        _4points = 0, 0, iw+dw, 0, iw+2*dw, ih, dw,ih
        img2 = crop_img_4points(img1,_4points)
        return img2 
   
    return img

def crop_img_4points(image, _4points):
    x1,y1,x2,y2,x3,y3,x4,y4 = _4points
    pts1 = np.float32([[x1,y1], [x2,y2], [x3,y3], [x4, y4]])
    width, height = calc_4points_shape(_4points)
    pts2 = np.float32([[0, 0], [width, 0], [width, height], [0, height]])
    M = cv2.getPerspectiveTransform(pts1, pts2)
    cropped = cv2.warpPerspective(image, M, (int(width), int(height)))

    return cropped

def calc_4points_shape(_4points):
    x1,y1,x2,y2,x3,y3,x4,y4 = _4points
    w1 = math.sqrt(math.pow(y2-y1, 2) + math.pow(x2-x1, 2))
    h1 = math.sqrt(math.pow(y3-y2, 2) + math.pow(x3-x2, 2))
    w2 = math.sqrt(math.pow(y4-y3, 2) + math.pow(x4-x3, 2))
    h2 = math.sqrt(math.pow(y4-y1, 2) + math.pow(x4-x1, 2))
    
    return (w1 + w2) / 2, (h1 + h2) / 2