def crnnRec(im, boxes, leftAdjust=False, rightAdjust=False, alph=0.2, f=1.0):
"""
crnn模型，ocr识别
@@model,
@@converter,
@@im:Array
@@text_recs:text box
@@ifIm:是否输出box对应的img
"""
results = []
im = Image.fromarray(im)
for index, box in enumerate(boxes):
degree, w, h, cx, cy = solve(box)
partImg, newW, newH = rotate_cut_img(im, degree, box, w, h, leftAdjust,
rightAdjust, alph)
newBox = xy_rotate_box(cx, cy, newW, newH, degree)
# partImg_ = partImg.convert('L')
# simPred = crnnOcr(partImg_)##识别的文本
# if simPred.strip()!=u'':
# results.append({'cx':cx*f,'cy':cy*f,'text':simPred,'w':newW*f,'h':newH*f,'degree':degree*180.0/np.pi})
results.append({
'cx': cx * f,
'cy': cy * f,
'text': '',
'w': newW * f,
'h': newH * f,
'degree': degree * 180.0 / np.pi
})
return results
def plot_boxes(img, angle, result, color=(0, 0, 0)):
tmp = np.array(img)
c = color
w, h = img.size
thick = int((h + w) / 300)
i = 0
if angle in [90, 270]:
imgW, imgH = img.size[::-1]
else:
imgW, imgH = img.size
for line in result:
cx = line['cx']
cy = line['cy']
degree = line['angle']
w = line['w']
h = line['h']
x1, y1, x2, y2, x3, y3, x4, y4 = xy_rotate_box(cx, cy, w, h, degree)
x1, y1, x2, y2, x3, y3, x4, y4 = box_rotate(
[x1, y1, x2, y2, x3, y3, x4, y4],
angle=(360 - angle) % 360,
imgH=imgH,
imgW=imgW)
cx = np.mean([x1, x2, x3, x4])
cy = np.mean([y1, y2, y3, y4])
cv2.line(tmp, (int(x1), int(y1)), (int(x2), int(y2)), c, 1)
cv2.line(tmp, (int(x2), int(y2)), (int(x3), int(y3)), c, 1)
cv2.line(tmp, (int(x3), int(y3)), (int(x4), int(y4)), c, 1)
cv2.line(tmp, (int(x4), int(y4)), (int(x1), int(y1)), c, 1)
mess = str(i)
cv2.putText(tmp, mess, (int(cx), int(cy)), 0, 1e-3 * h, c, thick // 2)
i += 1
return Image.fromarray(tmp)
def plot_boxes(img, result, angle=0, color=(0, 0, 0)):
tmp = np.array(img)
c = color
h, w = img.shape[:2]
thick = int((h + w) / 300)
i = 0
if angle in [90, 270]:
imgW, imgH = img.shape[:2]
else:
imgH, imgW = img.shape[:2]
for line in result:
cx = line['cx']
cy = line['cy']
degree = line['degree']
w = line['w']
h = line['h']
x1, y1, x2, y2, x3, y3, x4, y4 = xy_rotate_box(cx, cy, w, h,
degree / 180 * np.pi)
x1, y1, x2, y2, x3, y3, x4, y4 = box_rotate(
[x1, y1, x2, y2, x3, y3, x4, y4],
angle=(360 - angle) % 360,
imgH=imgH,
imgW=imgW)
cx = np.mean([x1, x2, x3, x4])
cy = np.mean([y1, y2, y3, y4])
cv2.line(tmp, (int(x1), int(y1)), (int(x2), int(y2)), c, 1)
cv2.line(tmp, (int(x2), int(y2)), (int(x3), int(y3)), c, 1)
cv2.line(tmp, (int(x3), int(y3)), (int(x4), int(y4)), c, 1)
cv2.line(tmp, (int(x4), int(y4)), (int(x1), int(y1)), c, 1)
mess = str(i)
cv2.putText(tmp, mess, (int(cx), int(cy)), 0, 1e-3 * h, c, thick // 2)
i += 1
_image = Image.fromarray(tmp).convert('RGB')
_image.save('test/result.png')
return _image
def api_root():
app.logger.info(PROJECT_HOME)
if request.method == 'POST' and request.files['image']:
starttime = time.time()
img = request.files['image']
img_cv = cv2.imdecode(np.fromstring(img.read(), np.uint8),
cv2.IMREAD_COLOR)
boxes, rec_res = text_sys(img_cv)
for box in boxes:
xy_sum = np.sum(box, axis=0) / 4.0
cx = xy_sum[0]
cy = xy_sum[1]
degree = np.arcsin(
(box[1][1] - box[0][1]) / (box[1][0] - box[0][0]))
w = abs(box[0][0] - box[1][0])
h = abs(box[0][1] - box[3][1])
x1, y1, x2, y2, x3, y3, x4, y4 = xy_rotate_box(
cx, cy, w, h, degree / 180 * np.pi)
box[0][0] = x1
box[0][1] = y1
box[1][0] = x2
box[1][1] = y2
box[2][0] = x3
box[2][1] = y3
box[3][0] = x4
box[3][1] = y4
assorted_results = [{
'box': boxes[i],
'txt': rec_res[i][0]
} for i in range(len(rec_res))]
res = trainTicket.trainTicket(assorted_results, img=img_cv)
res = res.res
elapse = time.time() - starttime
app.logger.info("Predict time : %.3fs" % elapse)
return json.dumps(res, ensure_ascii=False)
else:
return "Where is the image?"