def OCR(filepath, OCRtarget):
global frcnn_pro
global OCR_pro
global select_pro
image_unit.convert_png2jpg(filepath, saveDir)
image_unit.add_side(saveDir)
frcnn_len = len(os.listdir(saveDir))
#frcnn_detect predict
sess, net = frcnn.init(0, 0, OCRtarget)
for count, x in enumerate(os.listdir(saveDir)):
results = frcnn.detect(sess, net, os.path.join(saveDir, x))
#xmin,ymin,xmax,ymax = result[0]*width, result[1]*height, result[2]*width, result[3]*height
#object_num = result[4] prob = result[5] object_name = result[6]
if len(results) != 0:
image_unit.frcnn_crop(os.path.join(saveDir, x), results, splitDir,
charDir)
frcnn_pro = float(count + 1) / float(frcnn_len) * 100.0
sess.close()
ch_image_set, en_image_set = OCR_unit.load_charimg(charDir, img_size=64)
image_set = ch_image_set + en_image_set
text = {}
val_dict, idx_dict = OCR_unit.detect(OCRmodel, 3700, image_set)
text = OCR_unit.predict_text(val_dict, idx_dict, OCRlabel, text)
return OCR_unit.select_target(text, OCRtarget, saveDir)
def segment_frcnn():
image_unit.convert_png2jpg(pngdir, savedir)
image_unit.add_side(savedir)
#frcnn_detect predict
sess, net = frcnn.init(0, 0)
for x in os.listdir(savedir):
results = frcnn.detect(sess, net, savedir + x)
#xmin,ymin,xmax,ymax = result[0]*width, result[1]*height, result[2]*width, result[3]*height
#object_num = result[4] prob = result[5] object_name = result[6]
if len(results) != 0:
image_unit.frcnn_crop(savedir + x, results, splitdir, chardir)
sess.close()
def OCR(filepath):
image_unit.convert_png2jpg(filepath, saveDir)
image_unit.add_side(saveDir)
#frcnn_detect predict
sess, net = frcnn.init(0,0)
for x in os.listdir(saveDir):
results = frcnn.detect(sess, net, os.path.join(saveDir, x))
#xmin,ymin,xmax,ymax = result[0]*width, result[1]*height, result[2]*width, result[3]*height
#object_num = result[4] prob = result[5] object_name = result[6]
if len(results) != 0:
image_unit.frcnn_crop(os.path.join(saveDir, x), results, splitDir, charDir)
sess.close()
ch_image_set, en_image_set = OCR_unit.load_charimg(charDir, img_size = 64)
image_set = ch_image_set + en_image_set
text = {}
val_dict, idx_dict = OCR_unit.detect(OCRmodel, 3700, image_set)
text = OCR_unit.predict_text(val_dict, idx_dict, OCRlabel, text)
return OCR_unit.select_target(text, [u'名称',u'注册号'])
def segment():
image_unit.convert_png2jpg(pngdir, savedir)
image_unit.add_side(savedir)
detection = ssd.CaffeDetection(gpu_id, model_def, model_weights,
image_resize, labelmap_file)
#ssd_detect predict
for x in os.listdir(savedir):
result = detection.detect(savedir + x)
#xmin,ymin,xmax,ymax = result[0]*width, result[1]*height, result[2]*width, result[3]*height
#object_num = result[4] prob = result[5] object_name = result[6]
if len(result) != 0:
image_unit.ssd_crop(savedir + x, result, splitdir)
#segmentation char
if os.path.exists(chardir):
shutil.rmtree(chardir)
os.mkdir(chardir)
os.mkdir(os.path.join(chardir, 'firm_num'))
os.mkdir(os.path.join(chardir, 'firm_name'))
for x in os.listdir(os.path.join(splitdir, 'firm_num')):
id = x
img_path = os.path.join(splitdir, 'firm_num', id, 'num.jpg')
char_path = os.path.join(chardir, 'firm_num', id)
if os.path.exists(img_path):
os.mkdir(char_path)
os.system('/home/zhangzm/win3/ch_rec_ssd/test/text_split ' + img_path +
' ' + char_path)
if os.path.exists(char_path):
image_unit.num_handle(char_path + '/')
for x in os.listdir(os.path.join(splitdir, 'firm_name')):
id = x
img_path = os.path.join(splitdir, 'firm_name', id, 'name.jpg')
char_path = os.path.join(chardir, 'firm_name', id)
if os.path.exists(img_path):
os.mkdir(char_path)
os.system('/home/zhangzm/win3/ch_rec_ssd/test/text_split ' + img_path +
r' ' + char_path)
if os.path.exists(char_path):
image_unit.cn_handle(char_path + '/')