def final_steps(xml_total, proxy_total):
if int(xml_total) > 0:
xml_c.create_xml(xml_total)
if int(proxy_total) > 0:
gp.get_proxy(proxy_total)
complete_msg = f"{'='*25} SCRIPT COMPLETE {'='*25}"
logger.info(complete_msg)
def upload_image():
if request.method == 'POST':
# check if the post request has the file part
if 'file' not in request.files:
return redirect(url_for("root"))
file = request.files['file']
ip = request.remote_addr
######################################
# get the value of multi_label_open from front page
# multi_label_open = request.files['file']
# False : indicate return multi label
# True : indicate return '1' default is True
######################################
# False : indicate return multi label
# True : indicate return '1' default is True
signal = str(request.form['identified'])
prediction_result = ''
top3_dic_names = {}
# if user does not select file, browser also submit a empty part without filename
if file.filename == '':
return redirect(url_for("root"))
if file and allowed_file(file.filename):
filestr = request.files['file'].read()
logging.info(
'From ' + ip + ' -> Upload image ' + file.filename +
' with size of {} kb'.format(len(filestr) / float(1000.0)))
try:
# convert string data to numpy array
npimg = numpy.fromstring(filestr, numpy.uint8)
# convert numpy array to image
img = cv2.imdecode(npimg, cv2.IMREAD_COLOR)
imagename = secure_filename(file.filename).lower() # '7.jpeg'
imageshape = img.shape # (1440, 2560, 3)
except:
print('Corrupt Image:', file.filename)
logging.info('Corrupt Image:' + file.filename)
return 'error'
# filename : is image path such as 'static/img_pool/7.jpeg'
# filename = os.path.join("static/img_pool", secure_filename(file.filename).lower())
start = time.clock()
executor = APPEntity()
if signal == '0':
all_coordinates = executor.predict_obj(img)
margin = 15
for arry in all_coordinates:
arry[-4] -= margin
arry[-3] -= margin
arry[-2] += margin
arry[-1] += margin
'''create xml'''
prediction_result = create_xml.create_xml(
imagename, imageshape, all_coordinates, {})
elif signal == '1':
all_coordinates = executor.predict_obj(img)
if len(all_coordinates) != 0:
margin = 10
for arry in all_coordinates:
arry[-4] -= margin
arry[-3] -= margin
arry[-2] += margin
arry[-1] += margin
all_coordinates, top3_dic_names = executor.predict_face(
img, all_coordinates)
prediction_result = create_xml.create_xml(
imagename, imageshape, all_coordinates, top3_dic_names)
elif signal == '2':
all_coordinates = json.loads(request.form.get('coordinates'))
all_coordinates_trans = []
if len(all_coordinates) != 0:
for arry in all_coordinates:
arry_trans = [
arry['Id'], arry['X'], arry['Y'],
arry['X'] + arry['W'], arry['Y'] + arry['H']
]
all_coordinates_trans.append(arry_trans)
all_coordinates, top3_dic_names = executor.predreict_back(
img, all_coordinates_trans)
prediction_result = create_xml.create_xml(imagename,
imageshape,
all_coordinates,
top3_dic_names,
need_id=True)
elif signal == '3':
_, all_coordinates_face = demo.demo(sess, net, img)
top3_dic_names_face = {}
if len(all_coordinates_face) != 0:
margin = 10
for arry in all_coordinates_face:
arry[-4] -= margin
arry[-3] -= margin
arry[-2] += margin
arry[-1] += margin
all_coordinates_face, top3_dic_names_face = executor.predict_face(
img, all_coordinates_face)
prediction_result_face = create_xml.create_xml(
imagename, imageshape, all_coordinates_face,
top3_dic_names_face)
all_coordinates = json.loads(request.form.get('coordinates'))
all_coordinates_trans = []
top3_dic_names_back = {}
all_coordinates_back = []
if len(all_coordinates) != 0:
for arry in all_coordinates:
arry_trans = [
arry['Id'], arry['X'], arry['Y'],
arry['X'] + arry['W'], arry['Y'] + arry['H']
]
all_coordinates_trans.append(arry_trans)
all_coordinates_back, top3_dic_names_back = executor.predict_back(
img, all_coordinates_trans)
prediction_result_back = create_xml.create_xml(
imagename,
imageshape,
all_coordinates_back,
top3_dic_names_back,
need_id=True)
prediction_result = prediction_result_face + prediction_result_back
elif signal == '4':
recognized_class, score = executor.predict_antnet(img)
prediction_result = recognized_class + "," + str(score)
end = time.clock()
logging.info('From ' + ip + ' -> Deal with ' + file.filename +
' using GPU : {} s'.format(end - start))
# return render_template("index.html", prediction_result=prediction_result)
return prediction_result
return redirect(url_for("root"))
while True:
if add_flag:
cv2.addWeighted(temp, 0.7, img, 0.3, 0, img)
if copy_flag:
temp = img.copy()
cv2.putText(img, put_text, (50, 50), cv2.FONT_HERSHEY_COMPLEX,
1, (255, 255, 255), 2)
cv2.imshow('image', img)
key = cv2.waitKey(50) & 0xFF
if key == ord('q') or key == 27:
break
elif key == ord('6'): # next image
item += 1
if len(text_class) > 0:
create_xml(image, nWidth, nHeigth, text_class,
positions, xml_file)
print('%s file saved to ---> %s file' %
(image, xml_name))
add_flag = False
copy_flag = False
text_class = []
positions = []
break
elif key == ord('4'): # former image
former_flag = 1
break
# else:
# break
if key == ord('q') or (key & 0xff) == 27:
break
def split_resized_img():
'''
data_intput_path
imgs & xmls
imgs_output_path
imgs
xmls_output_path
xmls
'''
print('begin split resized imgs')
data_intput_path = os.path.join(data_path,
data_type + '_resized_2000_imgs')
imgs_output_path = os.path.join(data_path,
data_type + '_splited_1200_imgs')
xmls_output_path = os.path.join(data_path,
data_type + '_splited_1200_xmls')
W_pieces = 2
H_pieces = 10
new_size = splited_width
if not os.path.exists(imgs_output_path):
os.makedirs(imgs_output_path)
if not os.path.exists(xmls_output_path):
os.makedirs(xmls_output_path)
imgs = [s for s in os.listdir(data_intput_path) if s[-4:] == '.jpg']
xmls = [s for s in os.listdir(data_intput_path) if s[-4:] == '.xml']
assert len(imgs) == len(xmls)
print(len(imgs))
for i in range(len(imgs)):
assert imgs[i][:-4] == xmls[i][:-4]
img_path = os.path.join(data_intput_path, imgs[i])
xml_path = os.path.join(data_intput_path, xmls[i])
tree = ET.parse(xml_path)
root = tree.getroot()
size = get_and_check(root, 'size', 1)
width = int(get_and_check(size, 'width', 1).text)
height = int(get_and_check(size, 'height', 1).text)
H_overlap = int((H_pieces * new_size - height) / (H_pieces - 1))
W_overlap = int((W_pieces * new_size - width) / (W_pieces - 1))
all_splited_imgs = []
for w in range(W_pieces):
for h in range(H_pieces):
#print(str(w)+" "+str(h))
file_name = xmls[i][:-4] + '_' + str(w) + '_' + str(h)
#print(file_name)
start_w = np.maximum(w * (new_size - W_overlap), 0)
end_w = np.minimum(start_w + new_size, width)
start_h = np.maximum(h * (new_size - H_overlap), 0)
end_h = np.minimum(start_h + new_size, height)
source_img = cv2.imread(img_path)
if (source_img.shape[0] >= source_img.shape[1]):
splited_img = source_img[start_h:end_h, start_w:end_w, :]
else:
splited_img = source_img[start_w:end_w, start_h:end_h, :]
img_details = {'img_data':splited_img, 'height':new_size, 'width':new_size, \
'bboxes':[], 'filename':file_name}
has_box_iou_too_small = False
for obj in get(root, 'object'):
category = get_and_check(obj, 'name', 1).text
category = category[:2]
if int(category) in commom_flaws:
bndbox = get_and_check(obj, 'bndbox', 1)
xmin = int(get_and_check(bndbox, 'xmin', 1).text)
ymin = int(get_and_check(bndbox, 'ymin', 1).text)
xmax = int(get_and_check(bndbox, 'xmax', 1).text)
ymax = int(get_and_check(bndbox, 'ymax', 1).text)
assert (xmax > xmin)
assert (ymax > ymin)
if ymin < start_h and ymax > end_h:
ymin = start_h
ymax = end_h
if xmin < start_w and xmax > end_w:
xmin = start_w
xmax = end_w
## bbox in splited img / bbox in raw img
IoU = iou([start_w, start_h, end_w, end_h],
[xmin, ymin, xmax, ymax])
if IoU < 0.8:
has_box_iou_too_small = True
break
xmin = xmin - start_w
ymin = ymin - start_h
xmax = xmax - start_w
ymax = ymax - start_h
img_details['bboxes'].append({
'class': category,
'x1': xmin,
'y1': ymin,
'x2': xmax,
'y2': ymax
})
if (has_box_iou_too_small):
continue
if (len(img_details['bboxes']) > 0):
all_splited_imgs.append(img_details)
else:
continue
if (len(all_splited_imgs) > 3):
np.random.shuffle(all_splited_imgs)
all_splited_imgs = all_splited_imgs[:3]
for i in range(len(all_splited_imgs)):
img_detail = all_splited_imgs[i]
splited_img_data = img_detail['img_data']
create_xml(img_detail, xmls_output_path)
'''
xml_path = os.path.join(xmls_output_path, img_detail['filename']+'.xml')
tree = ET.parse(xml_path)
root = tree.getroot()
for obj in get(root, 'object'):
category = get_and_check(obj, 'name', 1).text
category = category[:2]
assert (int(category) > 60) & (int(category) < 90)
bndbox = get_and_check(obj, 'bndbox', 1)
xmin = int(int(get_and_check(bndbox, 'xmin', 1).text))
ymin = int(int(get_and_check(bndbox, 'ymin', 1).text))
xmax = int(int(get_and_check(bndbox, 'xmax', 1).text))
ymax = int(int(get_and_check(bndbox, 'ymax', 1).text))
assert(xmax > xmin)
assert(ymax > ymin)
cv2.rectangle(splited_img_data,(xmin, ymin), (xmax, ymax), (0,255,255),1)
textLabel = category
textOrg = (xmin, ymin)
cv2.putText(splited_img_data, textLabel, textOrg, cv2.FONT_HERSHEY_DUPLEX, 1, (0, 0, 0), 1)
'''
Image.fromarray(splited_img_data).save(imgs_output_path + sep +
img_detail['filename'] +
'.jpg')
print('split resized imgs end')
def resize_raw_imgs():
'''
new_W
图片进行缩小后的宽度，高度同比缩小
data_input_path
imgs & xmls
data_output_path
imgs & xmls
'''
print('begin resize raw imgs')
new_W = resized_width
data_input_path = os.path.join(data_path, data_type + '_raw_imgs')
data_output_path = os.path.join(data_path,
data_type + '_resized_2000_imgs')
if not os.path.exists(data_output_path):
os.makedirs(data_output_path)
imgs = [s for s in os.listdir(data_input_path) if s[-4:] == '.jpg']
xmls = [s for s in os.listdir(data_input_path) if s[-4:] == '.xml']
assert len(imgs) == len(xmls)
print(data_input_path + " " + str(len(imgs)))
for i in range(len(imgs)):
assert imgs[i][:-4] == xmls[i][:-4]
img_path = os.path.join(data_input_path, imgs[i])
xml_path = os.path.join(data_input_path, xmls[i])
## resize img
tree = ET.parse(xml_path)
root = tree.getroot()
size = get_and_check(root, 'size', 1)
width = int(get_and_check(size, 'width', 1).text)
height = int(get_and_check(size, 'height', 1).text)
if width <= 0 | height <= 0:
continue
ratio = float(new_W) / width
new_H = int(height * ratio)
raw_img = cv2.imread(img_path)
resized_img = cv2.resize(raw_img, (new_W, new_H))
img_details = {'height':new_H, 'width':new_W, \
'bboxes':[], 'filename':xmls[i][:-4]}
## resize xml
for obj in get(root, 'object'):
category = get_and_check(obj, 'name', 1).text
category = category[:2]
assert (int(category) > 60) & (int(category) < 90)
bndbox = get_and_check(obj, 'bndbox', 1)
xmin = int(int(get_and_check(bndbox, 'xmin', 1).text) * ratio)
ymin = int(int(get_and_check(bndbox, 'ymin', 1).text) * ratio)
xmax = int(int(get_and_check(bndbox, 'xmax', 1).text) * ratio)
ymax = int(int(get_and_check(bndbox, 'ymax', 1).text) * ratio)
assert (xmax > xmin)
assert (ymax > ymin)
'''
cv2.rectangle(raw_img,(xmin, ymin), (xmax, ymax), (0,0,255),2)
textLabel = category
textOrg = (xmin, ymin)
cv2.putText(raw_img, textLabel, textOrg, cv2.FONT_HERSHEY_DUPLEX, 1, (0, 0, 0), 1)
Image.fromarray(raw_img).save(data_output_path+sep+imgs[i])
'''
img_details['bboxes'].append({
'class': category,
'x1': xmin,
'y1': ymin,
'x2': xmax,
'y2': ymax
})
create_xml(img_details, data_output_path)
'''
xml_path = os.path.join(val_resized_imgs, xmls[i])
tree = ET.parse(xml_path)
root = tree.getroot()
for obj in get(root, 'object'):
category = get_and_check(obj, 'name', 1).text
category = category[:2]
assert (int(category) > 60) & (int(category) < 90)
bndbox = get_and_check(obj, 'bndbox', 1)
xmin = int(int(get_and_check(bndbox, 'xmin', 1).text))
ymin = int(int(get_and_check(bndbox, 'ymin', 1).text))
xmax = int(int(get_and_check(bndbox, 'xmax', 1).text))
ymax = int(int(get_and_check(bndbox, 'ymax', 1).text))
assert(xmax > xmin)
assert(ymax > ymin)
cv2.rectangle(resized_img,(xmin, ymin), (xmax, ymax), (0,0,255),2)
textLabel = category
textOrg = (xmin, ymin)
cv2.putText(resized_img, textLabel, textOrg, cv2.FONT_HERSHEY_DUPLEX, 1, (0, 0, 0), 1)
'''
Image.fromarray(resized_img).save(data_output_path + sep + imgs[i])
print('resize raw imgs end')
"").replace(";", "")
values = values.replace("(", "").replace(")", "")
values = values.replace("&", "").split(",")
pin_map[prev] += [
pin.strip() for pin in values if pin.strip() != ""
]
"""
# Show Port-to-Pin Mappings
for k in pin_map:
print(k, pin_map[k])
"""
with open(lbr_template, 'r') as f:
t = f.readlines()
template = "".join(t)
# Find name of device
matches = re.findall(r'(\bentity\b\ +\w+\b is\b)', s, re.IGNORECASE)
device = matches[0][6:-3].strip()
# Sort and create HTML
l_grp, r_grp, l_count, r_count = sort_ports(groups)
lbr = cx.create_xml(l_grp, r_grp, l_count, r_count, device, template,
pin_map)
# Write out to file
lbr_nm = "bsdl_files/" + device + ".lbr"
lbr_file = open(lbr_nm, 'w')
lbr_file.write(lbr)
lbr_file.close()