def test(fileList, model):
# prepare
prep()
fileAm = len(fileList)
print('Start testing ......\n')
print('Total : %d' % fileAm )
i, j = 0, list(map(int, np.multiply( fileAm, np.multiply(0.1, range(1,11,1)) )))
# test every img
for file in fileList:
#file = trainPath + file
# print progress
i = i + 1
if i in j:
print( '\t{:.0%}'.format((j.index(i)+1)/10 ))
# get predict
predict = seg(file, model, trainPath)
# get ground truth
trueMask = rle_mask(file)
# evaluate predict
evaluate = evalue( predict, trueMask )
#show2(trueMask, 'true', predict, str(evaluate))
# store img,evaluate
storePredict(file.replace(trainPath,''), evaluate)
# compute accuracy
accuracy = calAccu()
return accuracy
def combine(file):
seg.seg(file)
hsv.hsv(file)
def test2(filename):
seg(filename, 0, os.getcwd(), os.getcwd()+'/test')
def my_link():
(wid, hei) = s.seg()
time.sleep(5)
return render_template('hello1.html', wid=wid, hei=hei)
que = "".join(q['question'].split())
q['answer'] = ans
q['question'] = que
if len(ans) == 0 or len(que) == 0:
continue
# ---
# if q['answer_span'][0] < 0 or q['answer_span'][1] < 0:
# # print(d)
# # print(article)
# # print(ans)
# article = d['article_title']+"。"+article
# answer_span = gen_answer_span(article, que, ans)
# print(answer_span, q['answer_span'])
# if answer_span[0]==-1:
# # 这里是抽象型回答，或标注错误的回答。没有办法生成answer_span，一共48个
# pass
# q['answer_span'] = answer_span
# ---
answer_span = gen_answer_span(article, que, ans)
q['answer_span'] = answer_span
json.dump(data, open(output_file, "w", encoding='utf-8'), ensure_ascii=False)
if __name__ == "__main__":
# 1. 预处理，修整原文不正常标点符号，保留原文的字符，将标题与原文拼接到一起
# 2. 生成字级别的answer_span，这样可以避免分词错误
# main1(input_file="../../train_data/question.json", output_file="../../train_data/question_with_span.json")
# 3. 对question_with_span 的 article question answer进行分词处理。
# 4. 针对字级别的answer_span 设计算法，生成词级别的answer_span
seg(input_file="../../train_data/question_with_span.json", output_file="../../train_data/question_with_span_seg.json")
default=0.4)
parser.add_argument("--file",
help="enter some quality limit",
nargs='?',
default='miro')
parser.add_argument("--unity",
help="enter some quality limit",
nargs='?',
default='D:/PROGFILES/unity/Editor/Unity.exe')
# nargs='?', default='D:/programs/unity/Editor/Unity.exe')
parser.add_argument("--detail",
help="enter some quality limit",
action='store_true',
default=False)
parser.add_argument("--sbsize", type=float, default=512)
args = parser.parse_args()
BLOCK_SIZE = args.blocksize
COLTHRESH = args.colthresh
TEXTHRESH = args.texthresh
FILE = args.file
DETAIL = args.detail
MULTI = args.multi
SBSIZE = args.sbsize
UNITY = args.unity
seg.seg(BLOCK_SIZE, COLTHRESH, TEXTHRESH, FILE, DETAIL, MULTI)
cb.create_bg(BLOCK_SIZE, COLTHRESH, TEXTHRESH, FILE, DETAIL, MULTI, SBSIZE)
#gen3d.gen_3d(BLOCK_SIZE, COLTHRESH, TEXTHRESH, FILE, DETAIL, MULTI, SBSIZE)
genunity.gen_unity(BLOCK_SIZE, COLTHRESH, TEXTHRESH, FILE, DETAIL, MULTI,
SBSIZE, UNITY)
def words(self):
return seg.seg(self.doc)
def handle(self, doc):
words = seg.seg(doc)
words = normal.filter_stop(words)
return words
# - run renameFiles.py setting your_path to the path to RIDGES-Fraktur
# - run go.py, but replace the path in the line with "### HERE ###" to the path to RIDGES-Fraktur
# - run checkLabels.py replacing letter_data with the path to the letter_data folder in RIDGES-Fraktur
# - repeat the above 3 steps for dta19
stopp = 0
# list of book folders in 'data'
os.chdir(folderpath) ### HERE ### !!!!!!!!!!!
folders = [x for x in os.listdir() if os.path.isdir(os.path.join(os.getcwd(), x))]
#folders.remove('letter_data')
#for foldername in folders: # for each book folder in 'data'
datapath = folderpath+folder # path to img/txt data for this book
try: # make folder in 'segmentation/segs' to store morph imgs for this book
os.mkdir(seg_dir+foldername)
except FileExistsError: pass
destpath = seg_dir+foldername # path to save segmented letter imgs for this book
for img in [x for x in os.listdir(datapath) if x[-3:] == 'png']: # for each line img in this book/folder
try: os.mkdir(datapath+'/temp') # to store word images
except FileExistsError: pass
wordspath = datapath+'/temp' # to store word images
wordSeg(img, datapath, wordspath) # get word segmented images
print('wordSeg', img, '~~~~~~~~~~~~~~~~~~~~~~~~~')
for wordImg in [x for x in os.listdir(wordspath) if x[-3:] == 'png']: # for each word img in this line
print('seg', wordImg)
seg(wordImg, 1, wordspath, destpath) # segment letters
shutil.rmtree(wordspath) # delete 'temp' folder
stopp += 1
if stopp > 8: break
labelData(destpath, your_path_here+'/fraktur/testdata/letter_data/') # separate letter imgs into folders
def test2(filename):
seg(filename, 0, os.getcwd(), os.getcwd() + '/test')
### execute ###
filename = 'white.png'
filename = 'rrr.png'
filename = 'uu.png'
filename = 'uuu_ſeiner.png'
filename = 'ess.png'
filename = 'ess1_Gelüſte.png'
# test1(filename)
seg(filename, 1, os.getcwd(), os.getcwd() + '/test')
# img, nimg, bin = preprocess(filename)
#
# _, contours, _ = cv2.findContours(bin.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
# contours = sorted(contours, key=lambda ctr: cv2.boundingRect(ctr)[0]) # sort contours in order
# contour = contours[0]
# cv2.imwrite('bin.png', bin)
#
# out = np.full_like(bin, 0) # create array of 0s (blacks)
# for c in contours:
# contour = c
# ones = np.ones_like(bin) # create array of 1s (almost blacks)
# mask = cv2.drawContours(ones, [contour], 0, 0, -1) # color contour area as 0
# out[mask == 0] = 255 # where mask is 0, change out value to img value
# x, y, w, h = cv2.boundingRect(contour) # get bounding box for cropping