def main():
pp = pprint.PrettyPrinter(indent=4)
img_path = args.img_path
print("imageee path",img_path)
label_path = args.label_path
img_type = args.img_type
datasets = args.datasets
cls_list = args.cls_list_file
result = None
data = None
if datasets == "COCO":
coco = COCO()
result, data = coco.parse(label_path)
elif datasets == "VOC":
voc = VOC()
result, data = voc.parse(label_path)
elif datasets == "UDACITY":
udacity = UDACITY()
result, data = udacity.parse(label_path, img_path)
elif datasets == "KITTI":
kitti = KITTI()
result, data = kitti.parse(label_path, img_path, img_type=img_type)
elif datasets == "YOLO":
yolo =YOLO(os.path.abspath(cls_list))
result, data = yolo.parse(label_path, img_path, img_type=img_type)
if result is True:
for key in data:
filepath = "".join([img_path, key, img_type])
im = Image.open(filepath)
draw = ImageDraw.Draw(im)
print("data['{}']: ".format(key), end="")
pp.pprint(data[key])
print("num_object : {}".format(data[key]["objects"]["num_obj"]))
for idx in range(0, int(data[key]["objects"]["num_obj"])):
print("idx {}, name : {}, bndbox :{}".format(idx, data[key]["objects"][str(idx)]["name"], data[key]["objects"][str(idx)]["bndbox"]))
x0 = data[key]["objects"][str(idx)]["bndbox"]["xmin"]
y0 = data[key]["objects"][str(idx)]["bndbox"]["ymin"]
x1 = data[key]["objects"][str(idx)]["bndbox"]["xmax"]
y1 = data[key]["objects"][str(idx)]["bndbox"]["ymax"]
draw.rectangle(((x0,y0), (x1,y1)), outline='#00ff88')
draw.text((x0,y0), data[key]["objects"][str(idx)]["name"])
del draw
print("===============================================================================================\n\n")
plt.imshow(im)
plt.show()
plt.clf()
im.close()
else:
print("return value : {}, msg : {}, args: {}".format(result, data, args))
def main(config):
if config["datasets"] == "VOC":
voc = VOC()
yolo = YOLO(os.path.abspath(config["cls_list"]))
print('parsing...')
flag, data = voc.parse(config["label"])
if flag == True:
print('parsing succeeded')
flag, data = yolo.generate(data)
if flag == True:
print('saving results')
if not os.path.exists(config["output_path"]):
os.makedirs(config["output_path"])
flag, data = yolo.save(data, config["output_path"],
config["img_path"], config["img_type"],
config["manifest_path"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("YOLO Generating Result : {}, msg : {}".format(
flag, data))
else:
print("VOC Parsing Result : {}, msg : {}".format(flag, data))
def main(config):
if config["datasets"] == "VOC":
voc = VOC()
yolo = YOLO(os.path.abspath(config["cls_list"]))
flag, data = voc.parse(config["label"])
if flag == True:
flag, data = yolo.generate(data)
if flag == True:
flag, data = yolo.save(data, config["output_path"],
config["img_path"], config["img_type"],
config["manipast_path"],
config["train"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("YOLO Generating Result : {}, msg : {}".format(
flag, data))
else:
print("VOC Parsing Result : {}, msg : {}".format(flag, data))
else:
print("Unknown Datasets")
def main(config):
if config["datasets"] == "VOC":
voc = VOC()
yolo = YOLO(os.path.abspath(config["cls_list"]))
flag, data = voc.parse(config["label"])
if flag == True:
flag, data = yolo.generate(data)
if flag == True:
flag, data = yolo.save(data, config["output_path"],
config["img_path"], config["img_type"],
config["manipast_path"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("YOLO Generating Result : {}, msg : {}".format(
flag, data))
else:
print("VOC Parsing Result : {}, msg : {}".format(flag, data))
elif config["datasets"] == "COCO":
coco = COCO()
yolo = YOLO(os.path.abspath(config["cls_list"]))
flag, data = coco.parse(config["label"])
if flag == True:
flag, data, data_annotation = yolo.generate(data)
if flag == True:
flag, data = yolo.save_annotation(data, data_annotation,
config["output_path"],
config["img_path"],
config["img_type"],
config["manipast_path"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("YOLO Generating Result : {}, msg : {}".format(
flag, data))
else:
print("COCO Parsing Result : {}, msg : {}".format(flag, data))
elif config["datasets"] == "UDACITY":
udacity = UDACITY()
yolo = YOLO(os.path.abspath(config["cls_list"]))
flag, data = udacity.parse(config["label"])
if flag == True:
flag, data = yolo.generate(data)
if flag == True:
flag, data = yolo.save(data, config["output_path"],
config["img_path"], config["img_type"],
config["manipast_path"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("UDACITY Generating Result : {}, msg : {}".format(
flag, data))
else:
print("COCO Parsing Result : {}, msg : {}".format(flag, data))
elif config["datasets"] == "KITTI":
kitti = KITTI()
yolo = YOLO(os.path.abspath(config["cls_list"]))
flag, data = kitti.parse(config["label"],
config["img_path"],
img_type=config["img_type"])
if flag == True:
flag, data = yolo.generate(data)
if flag == True:
flag, data = yolo.save(data, config["output_path"],
config["img_path"], config["img_type"],
config["manipast_path"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("YOLO Generating Result : {}, msg : {}".format(
flag, data))
else:
print("KITTI Parsing Result : {}, msg : {}".format(flag, data))
else:
print("Unkwon Datasets")
def generate_voc_annotations():
image_num = 0
images_train = []
xml_files = get_xml_files()
voc_format = VOC()
yolo_format = YOLO(os.path.abspath(os.path.join(data_yolo, 'penelope.name').replace("\\", "/")))
with tqdm(bar_format='{l_bar}{bar}{n_fmt}/{total_fmt} [{elapsed}<{remaining}]') as pbar:
for xml_file in xml_files:
root = ElementTree.parse(xml_file).getroot()
title = root.get('title')
pages = root.find('pages')
pbar.total = len(xml_files)
pbar.set_description("Labeling {0}".format(title), refresh=True)
# pbar.reset()
pbar.update()
for page in pages:
index_img = page.get('index').zfill(3)
width = page.get('width')
height = page.get('height')
texts = page.findall('text')
# Some page doesn't have texts
if not texts:
continue
image_path = os.path.join(images_dir, title, index_img + '.jpg').replace("\\", "/")
new_image_path = os.path.join(data_yolo_labels, str(image_num) + '.jpg').replace("\\", "/")
copyfile(image_path, new_image_path)
images_train.append(new_image_path + '\n')
voc_file = os.path.join(images_voc_dir, str(image_num) + '.xml').replace("\\", "/")
voc_root = ElementTree.Element("annotations")
ElementTree.SubElement(voc_root, "filename").text = "{}.jpg".format(image_num)
ElementTree.SubElement(voc_root, "folder").text = new_image_path
source = ElementTree.SubElement(voc_root, "source")
ElementTree.SubElement(source, "database").text = "Unknown"
size = ElementTree.SubElement(voc_root, "size")
ElementTree.SubElement(size, "width").text = str(width)
ElementTree.SubElement(size, "height").text = str(height)
ElementTree.SubElement(size, "depth").text = "3"
ElementTree.SubElement(voc_root, "segmented").text = str(0)
for each_text in texts:
# Draw rectangle around the text
# +/- 10 because some letters are output of the rectangle
x_min = int(each_text.get('xmin')) - 10
x_max = int(each_text.get('xmax')) + 10
y_min = int(each_text.get('ymin')) - 10
y_max = int(each_text.get('ymax')) + 10
text = each_text.text
if text is not None:
obj = ElementTree.SubElement(voc_root, "object")
ElementTree.SubElement(obj, "name").text = "text"
ElementTree.SubElement(obj, "pose").text = "Unspecified"
ElementTree.SubElement(obj, "truncated").text = str(0)
ElementTree.SubElement(obj, "difficult").text = str(0)
bbox = ElementTree.SubElement(obj, "bndbox")
ElementTree.SubElement(bbox, "xmin").text = str(x_min)
ElementTree.SubElement(bbox, "ymin").text = str(y_min)
ElementTree.SubElement(bbox, "xmax").text = str(x_max)
ElementTree.SubElement(bbox, "ymax").text = str(y_max)
# TEST PYTESSERACT
# RESULT : FAILED
# image_roi = cv_image[y_min:y_max, x_min:x_max]
# configuration = "-l jpn_vert --oem 1 --psm 5"
# pytext = pytesseract.image_to_string(image_roi, config=configuration).replace("\n", "").replace(" ", "")
# if text == pytext:
# print("It's equal {0} and {1}".format(text, pytext))
# else:
# print("It's NOT equal {0} and {1}".format(text, pytext))
# assert text == pytext, "Difference between {0} and {1}".format(text, pytext)
# For debugging
# cv2.rectangle(cv_image, (int(x_min), int(y_min)), (int(x_max), int(y_max)), (0, 255, 0), 2)
image_num = image_num + 1
xml_str = minidom.parseString(ElementTree.tostring(voc_root)).toprettyxml(indent=" ")
with open(voc_file, "w") as f:
f.write(xml_str)
with open(os.path.join(data_yolo, 'train.txt').replace("\\", "/"), 'w') as f:
f.writelines(images_train)
flag, data = voc_format.parse(images_voc_dir)
if flag:
flag, data = yolo_format.generate(data)
if flag:
flag, data = yolo_format.save(data, data_yolo_labels, data_yolo_labels, '.jpg', data_yolo_labels)
if not flag:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("YOLO Generating Result : {}, msg : {}".format(flag, data))
def main(config):
if config["datasets"] == "VOC":
voc = VOC()
yolo = YOLO(os.path.abspath(config["cls_list"]))
flag, data = voc.parse(config["label"])
if flag == True:
flag, data = yolo.generate(data)
if flag == True:
flag, data = yolo.save(data, config["output_path"],
config["img_path"], config["img_type"],
config["manifest_path"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("YOLO Generating Result : {}, msg : {}".format(
flag, data))
else:
print("VOC Parsing Result : {}, msg : {}".format(flag, data))
elif config["datasets"] == "COCO":
coco = COCO()
keep = {
"person", "bicycle", "car", "motorcycle", "bus", "train", "truck"
}
flag, data, cls_hierarchy = coco.parse(config["label"],
config["img_path"],
keep=keep)
data = sampleDataset(data, config["num_samples"], keep, config["seed"])
yolo = YOLO(os.path.abspath(config["cls_list"]),
cls_hierarchy=cls_hierarchy)
if flag == True:
flag, data = yolo.generate(data)
if flag == True:
flag, data = yolo.save(data, config["output_path"],
config["img_path"], config["img_type"],
config["manifest_path"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("YOLO Generating Result : {}, msg : {}".format(
flag, data))
else:
print("COCO Parsing Result : {}, msg : {}".format(flag, data))
elif config["datasets"] == "UDACITY":
udacity = UDACITY()
yolo = YOLO(os.path.abspath(config["cls_list"]))
flag, data = udacity.parse(config["label"])
if flag == True:
flag, data = yolo.generate(data)
if flag == True:
flag, data = yolo.save(data, config["output_path"],
config["img_path"], config["img_type"],
config["manifest_path"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("UDACITY Generating Result : {}, msg : {}".format(
flag, data))
else:
print("COCO Parsing Result : {}, msg : {}".format(flag, data))
elif config["datasets"] == "KITTI":
kitti = KITTI()
yolo = YOLO(os.path.abspath(config["cls_list"]))
flag, data = kitti.parse(config["label"],
config["img_path"],
img_type=config["img_type"])
if flag == True:
flag, data = yolo.generate(data)
if flag == True:
flag, data = yolo.save(data, config["output_path"],
config["img_path"], config["img_type"],
config["manifest_path"])
if flag == False:
print("Saving Result : {}, msg : {}".format(flag, data))
else:
print("YOLO Generating Result : {}, msg : {}".format(
flag, data))
else:
print("KITTI Parsing Result : {}, msg : {}".format(flag, data))
else:
print("Unkwon Datasets")