def download_quickdraw_414k(output_folder="./", remove_sourcefile=True):
print(torchsketch())
time.sleep(2)
print_banner("Downloading started......")
try:
gdown.download(
"https://drive.google.com/uc?id=1q933KpmJGkfStgIbwMfgfls1_1ZJVFyd",
os.path.join(output_folder, "picture_files.tar.gz"),
quiet=False)
gdown.download(
"https://drive.google.com/uc?id=1Vrf1ouhtWYJp4XKa6jestLGY3aVlxfLC",
os.path.join(output_folder, "coordinate_files.tar.gz"),
quiet=False)
except Exception as e:
print(e)
return
print("\n")
print_banner(
"quickdraw_414k is downloaded {}!".format('\033[32;1m' +
'succesfully' + '\033[0m'))
assert md5sum(os.path.join(
output_folder, "picture_files.tar.gz")) == MD5SUM_TABLE[
"picture_files.tar.gz"], 'picture_files.tar.gz md5 checksum error'
assert md5sum(
os.path.join(output_folder, "coordinate_files.tar.gz")
) == MD5SUM_TABLE[
"coordinate_files.tar.gz"], 'coordinate_files.tar.gz md5 checksum error'
print_banner("Md5 checksum passed {}!".format('\033[32;1m' +
'succesfully' + '\033[0m'))
print_banner("Extracting started......")
try:
extract_files(file_name=os.path.join(output_folder,
"picture_files.tar.gz"),
output_folder=output_folder,
remove_sourcefile=remove_sourcefile)
extract_files(file_name=os.path.join(output_folder,
"coordinate_files.tar.gz"),
output_folder=output_folder,
remove_sourcefile=remove_sourcefile)
except Exception as e:
print(e)
return
print_banner(
"quickdraw_414k is extracted {}!".format('\033[32;1m' + 'succesfully' +
'\033[0m'))
print(torchsketch())
def randomly_remove_strokes_4_svg(svg_url,
output_folder,
remove_ratio=0.3,
verbose=True):
if remove_ratio <= 0.0 or remove_ratio >= 1.0:
print(
"The input parameter of remove_ratio should be defined in (0.0, 1.0)."
)
print("Please choose a proper remove_ratio value, and retry this API.")
print(torchsketch())
return 0
doc = minidom.parse(svg_url)
paths = doc.getElementsByTagName("path")
assert type(paths) == xml.dom.minicompat.NodeList
stroke_count = len(paths)
if verbose == True:
print("{} has {} storkes.".format(svg_url, stroke_count))
reset_dir(output_folder)
svg_filename = svg_url.split("/")[-1][:-4]
seed = int(time.time())
np.random.seed(seed)
index_list = np.arange(stroke_count)
np.random.shuffle(index_list)
random_list = index_list[:int(stroke_count * remove_ratio)]
try:
for i in random_list:
doc.getElementsByTagName("path")[int(i)].setAttribute(
"visibility", "hidden")
output = open(
output_folder + "/" + svg_filename + "_" +
"randomly_partially_removed" + ".svg", "w")
doc.writexml(output)
output.close()
except Exception as e:
print(e)
if verbose == True:
print("Details are as follows.")
print("{} strokes have been removed RANDOMLY.".format(
int(stroke_count * remove_ratio)))
print("The produced sketch is stored in {}".format(output_folder))
print(torchsketch())
return stroke_count
def download_tu_berlin(output_folder="./", remove_sourcefile=True):
print(torchsketch())
time.sleep(2)
print_banner("Downloading started......")
try:
wget.download(
"http://cybertron.cg.tu-berlin.de/eitz/projects/classifysketch/sketches_svg.zip",
out=os.path.join(output_folder, "sketches_svg.zip"))
wget.download(
"http://cybertron.cg.tu-berlin.de/eitz/projects/classifysketch/sketches_png.zip",
out=os.path.join(output_folder, "sketches_png.zip"))
except Exception as e:
print(e)
return
print("\n")
print_banner(
"TU-Berlin is downloaded {}!".format('\033[32;1m' + 'succesfully' +
'\033[0m'))
assert md5sum(os.path.join(
output_folder, "sketches_svg.zip")) == MD5SUM_TABLE[
"sketches_svg.zip"], 'sketches_svg.zip md5 checksum error'
assert md5sum(os.path.join(
output_folder, "sketches_png.zip")) == MD5SUM_TABLE[
"sketches_png.zip"], 'sketches_png.zip md5 checksum error'
print_banner("Md5 checksum passed {}!".format('\033[32;1m' +
'succesfully' + '\033[0m'))
print_banner("Extracting started......")
try:
extract_files(file_name=os.path.join(output_folder,
"sketches_svg.zip"),
output_folder=output_folder,
remove_sourcefile=remove_sourcefile)
extract_files(file_name=os.path.join(output_folder,
"sketches_png.zip"),
output_folder=output_folder,
remove_sourcefile=remove_sourcefile)
except Exception as e:
print(e)
return
print_banner(
"TU-Berlin is extracted {}!".format('\033[32;1m' + 'succesfully' +
'\033[0m'))
print(torchsketch())
def convert_svgs_2_pngs(svg_url_list, output_folder = None):
for svg_url in svg_url_list:
convert_svg_2_png(svg_url, output_folder)
print(torchsketch())
def convert_colors_4_svgs(svg_url_list, output_folder, colors = ["red", ], verbose = True):
for svg_url in svg_url_list:
convert_colors_4_svg(svg_url, output_folder, colors, verbose)
print(torchsketch())
def download_qmul_chair(output_folder="./", remove_sourcefile=True):
print(torchsketch())
time.sleep(2)
print_banner("Downloading started......")
try:
gdown.download(
"https://drive.google.com/uc?id=1xlBhTJQwtssi8oJHbYbqVer8ZmyEG3wq",
os.path.join(output_folder, "chairs.zip"),
quiet=False)
except Exception as e:
print(e)
return
print("\n")
print_banner(
"QMUL chair is downloaded {}!".format('\033[32;1m' + 'succesfully' +
'\033[0m'))
assert md5sum(
os.path.join(output_folder, "chairs.zip")
) == MD5SUM_TABLE["chairs.zip"], 'chairs.zip md5 checksum error'
print_banner("Md5 checksum passed {}!".format('\033[32;1m' +
'succesfully' + '\033[0m'))
print_banner("Extracting started......")
try:
extract_files(file_name=os.path.join(output_folder, "chairs.zip"),
output_folder=output_folder,
remove_sourcefile=remove_sourcefile)
except Exception as e:
print(e)
return
print_banner(
"QMUL chair is extracted {}!".format('\033[32;1m' + 'succesfully' +
'\033[0m'))
print(torchsketch())
def download_qmul_shoe(output_folder="./", remove_sourcefile=True):
print(torchsketch())
time.sleep(2)
print_banner("Downloading started......")
try:
gdown.download(
"https://drive.google.com/uc?id=1S9lHUzdgR9yIRuIAE0kTw_odWVzb0Q5l",
os.path.join(output_folder, "shoes.zip"),
quiet=False)
except Exception as e:
print(e)
return
print("\n")
print_banner(
"QMUL shoe is downloaded {}!".format('\033[32;1m' + 'succesfully' +
'\033[0m'))
assert md5sum(
os.path.join(output_folder, "shoes.zip")
) == MD5SUM_TABLE["shoes.zip"], 'shoes.zip md5 checksum error'
print_banner("Md5 checksum passed {}!".format('\033[32;1m' +
'succesfully' + '\033[0m'))
print_banner("Extracting started......")
try:
extract_files(file_name=os.path.join(output_folder, "shoes.zip"),
output_folder=output_folder,
remove_sourcefile=remove_sourcefile)
except Exception as e:
print(e)
return
print_banner(
"QMUL shoe is extracted {}!".format('\033[32;1m' + 'succesfully' +
'\033[0m'))
print(torchsketch())
def convert_stroke_width_4_svgs(svg_url_list,
output_folder,
stroke_width=6,
verbose=True):
for svg_url in svg_url_list:
convert_stroke_width_4_svg(svg_url, output_folder, stroke_width,
verbose)
print(torchsketch())
def convert_colors_4_svg(svg_url,
output_folder,
colors=[
"red",
],
verbose=True):
doc = minidom.parse(svg_url)
paths = doc.getElementsByTagName("path")
assert type(paths) == xml.dom.minicompat.NodeList
stroke_count = len(paths)
if verbose == True:
print("{} has {} storkes.".format(svg_url, len(paths)))
if not os.path.exists(output_folder):
os.makedirs(output_folder)
produced_count = 0
svg_filename = svg_url.split("/")[-1][:-4]
for color in colors:
if color not in COLOR_TABLE.keys():
continue
try:
for i in range(stroke_count):
doc.getElementsByTagName("path")[i].setAttribute(
"stroke", color)
except Exception as e:
print(e)
continue
output = open(
output_folder + "/" + svg_filename + "_" + str(color) + ".svg",
"w")
doc.writexml(output)
output.close()
produced_count += 1
if verbose == True:
print("Details are as follows.")
print("{} colors have been inputted.".format(len(colors)))
print(
"{} colorful sketches have been produced.".format(produced_count))
print("{} colors can not been recognized.".format(
len(colors) - produced_count))
print(torchsketch())
return produced_count
def count_strokes_4_svg(svg_url, verbose=True):
doc = minidom.parse(svg_url)
path_strings = [
path.getAttribute('d') for path in doc.getElementsByTagName('path')
]
doc.unlink()
assert type(path_strings) == list
if verbose == True:
print("{} has {} storkes.".format(svg_url, len(path_strings)))
print(torchsketch())
return len(path_strings)
def randomly_convert_stroke_width_4_svg(svg_url, output_folder, stroke_width = 6, selected_strokes = 3, verbose = True):
doc = minidom.parse(svg_url)
paths = doc.getElementsByTagName("path")
assert type(paths) == xml.dom.minicompat.NodeList
stroke_count = len(paths)
if verbose == True:
print("{} has {} storkes.".format(svg_url, len(paths)))
if not os.path.exists(output_folder):
os.makedirs(output_folder)
svg_filename = svg_url.split("/")[-1][:-4]
seed = int(time.time())
np.random.seed(seed)
index_list = np.arange(stroke_count)
np.random.shuffle(index_list)
random_list = index_list[ : selected_strokes]
try:
for i in random_list:
doc.getElementsByTagName("path")[i].setAttribute("stroke-width", str(stroke_width))
except Exception as e:
print(e)
output = open(output_folder + "/" + svg_filename + "_randomly_selected_" + str(selected_strokes) + "_strokes_stroke_width_" + str(stroke_width) + ".svg", "w")
doc.writexml(output)
output.close()
if verbose == True:
print("Details are as follows.")
print("{} strokes have been selected randomly.".format(selected_strokes))
print("The stroke widths have been modified as {}.".format(stroke_width))
print(torchsketch())
return stroke_width
def convert_svg_2_accumulative_svgs(svg_url, output_folder, verbose=True):
doc = minidom.parse(svg_url)
paths = doc.getElementsByTagName("path")
assert type(paths) == xml.dom.minicompat.NodeList
stroke_count = len(paths)
if verbose == True:
print("{} has {} storkes.".format(svg_url, len(paths)))
reset_dir(output_folder)
svg_filename = svg_url.split("/")[-1][:-4]
try:
output = open(
output_folder + "/" + svg_filename + "_" +
str(stroke_count).zfill(6) + ".svg", "w")
doc.writexml(output)
output.close()
for i in range(stroke_count - 1, 0, -1):
doc.getElementsByTagName("path")[i].setAttribute(
"visibility", "hidden")
output = open(
output_folder + "/" + svg_filename + "_" + str(i).zfill(6) +
".svg", "w")
doc.writexml(output)
output.close()
except Exception as e:
print(e)
if verbose == True:
print("Details are as follows.")
print("{} svgs have been produced, and stored in {}.".format(
stroke_count, output_folder))
print(torchsketch())
return stroke_count
def convert_svg_2_gif(svg_url, output_folder, gif_fps=3, verbose=True):
convert_svg_2_accumulative_svgs(svg_url=svg_url,
output_folder=output_folder,
verbose=True)
if verbose == True:
print("Accumulative svgs have been produced.")
accumulative_svg_list = os.listdir(output_folder)
accumulative_svg_list.sort(key=lambda i: int(i[-10:-4]))
gif_frames = list()
for accum_svg in accumulative_svg_list:
accum_svg_url = os.path.join(output_folder, accum_svg)
converted_pdf_url = accum_svg_url[:-4] + ".pdf"
cairosvg.svg2pdf(url=accum_svg_url, write_to=converted_pdf_url)
gif_frames.append(
convert_from_path(pdf_path=converted_pdf_url, dpi=50)[0])
os.remove(accum_svg_url)
os.remove(converted_pdf_url)
imageio.mimsave(output_folder + "/" + svg_url.split("/")[-1][:-4] + ".gif",
gif_frames,
fps=gif_fps)
gif_frame_count = len(gif_frames)
if verbose == True:
print("Intermediate files have been well deleted.")
print("The achieved gif file has {} frames in total.".format(
gif_frame_count))
print(torchsketch())
return gif_frame_count
def download_sketchy(output_folder="./", remove_sourcefile=True):
print(torchsketch())
time.sleep(2)
print_banner("Downloading started......")
try:
gdown.download(
"https://drive.google.com/uc?id=0B7ISyeE8QtDdbUpYWV8tcFJlY2M",
os.path.join(output_folder, "sketches-06-04.7z"),
quiet=False)
gdown.download(
"https://drive.google.com/uc?id=0B7ISyeE8QtDdTjE1MG9Gcy1kSkE",
os.path.join(output_folder, "rendered_256x256.7z"),
quiet=False)
gdown.download(
"https://drive.google.com/uc?id=0B7ISyeE8QtDdaFhqeTZiNVBYZjA",
os.path.join(output_folder, "info-06-04.7z"),
quiet=False)
except Exception as e:
print(e)
return
print("\n")
print_banner("Sketchy is downloaded {}!".format('\033[32;1m' +
'succesfully' + '\033[0m'))
assert md5sum(os.path.join(
output_folder, "sketches-06-04.7z")) == MD5SUM_TABLE[
"sketches-06-04.7z"], 'sketches-06-04.7z md5 checksum error'
assert md5sum(os.path.join(
output_folder, "rendered_256x256.7z")) == MD5SUM_TABLE[
"rendered_256x256.7z"], 'rendered_256x256.7z md5 checksum error'
assert md5sum(
os.path.join(output_folder, "info-06-04.7z")
) == MD5SUM_TABLE["info-06-04.7z"], 'info-06-04.7z md5 checksum error'
print_banner("Md5 checksum passed {}!".format('\033[32;1m' +
'succesfully' + '\033[0m'))
print_banner("Extracting started......")
try:
extract_files(file_name=os.path.join(output_folder,
"sketches-06-04.7z"),
output_folder=output_folder,
remove_sourcefile=remove_sourcefile)
extract_files(file_name=os.path.join(output_folder,
"rendered_256x256.7z"),
output_folder=output_folder,
remove_sourcefile=remove_sourcefile)
extract_files(file_name=os.path.join(output_folder, "info-06-04.7z"),
output_folder=output_folder,
remove_sourcefile=remove_sourcefile)
except Exception as e:
print(e)
return
print_banner("Sketchy is extracted {}!".format('\033[32;1m' +
'succesfully' + '\033[0m'))
print(torchsketch())
def remove_strokes_4_svg(svg_url,
output_folder,
remove_ratio=0.3,
verbose=True):
if remove_ratio <= 0.0 or remove_ratio >= 1.0:
print(
"The input parameter of remove_ratio should be defined in (0.0, 1.0)."
)
print("Please choose a proper remove_ratio value, and retry this API.")
print(torchsketch())
return 0
doc = minidom.parse(svg_url)
paths = doc.getElementsByTagName("path")
assert type(paths) == xml.dom.minicompat.NodeList
stroke_count = len(paths)
stroke_length_array = [len(path.getAttribute('d')) for path in paths]
assert len(stroke_length_array) == stroke_count
stroke_length_array.sort()
threshold_value = stroke_length_array[int(stroke_count * remove_ratio)]
if verbose == True:
print("{} has {} storkes.".format(svg_url, stroke_count))
reset_dir(output_folder)
svg_filename = svg_url.split("/")[-1][:-4]
try:
for i in range(stroke_count):
if len(doc.getElementsByTagName("path")[i].getAttribute(
"d")) <= threshold_value:
doc.getElementsByTagName("path")[i].setAttribute(
"visibility", "hidden")
output = open(
output_folder + "/" + svg_filename + "_" + "partially_removed" +
".svg", "w")
doc.writexml(output)
output.close()
except Exception as e:
print(e)
if verbose == True:
print("Details are as follows.")
print("{} strokes have been removed.".format(
int(stroke_count * remove_ratio)))
print("The produced sketch is stored in {}".format(output_folder))
print(torchsketch())
return stroke_count
def mark_longest_strokes_4_svg(svg_url, output_folder, mark_longest = True, color = "red", verbose = True):
if color not in color_table.keys():
print("Please choose a recognizable color.")
return 0
doc = minidom.parse(svg_url)
paths = doc.getElementsByTagName("path")
assert type(paths) == xml.dom.minicompat.NodeList
stroke_count = len(paths)
stroke_length_array = [len(path.getAttribute('d')) for path in paths]
assert len(stroke_length_array) == stroke_count
stroke_length_array.sort()
if mark_longest:
threshold_value = stroke_length_array[-1]
else:
threshold_value = stroke_length_array[0]
if verbose == True:
print("{} has {} storkes.".format(svg_url, stroke_count))
reset_dir(output_folder)
svg_filename = svg_url.split("/")[-1][:-4]
marked_count = 0
try:
for i in range(stroke_count):
if len(doc.getElementsByTagName("path")[i].getAttribute("d")) == threshold_value:
doc.getElementsByTagName("path")[i].setAttribute("stroke", color)
marked_count += 1
if mark_longest:
output = open(output_folder + "/" + svg_filename + "_" + "marked_longest_stroke" + ".svg", "w")
else:
output = open(output_folder + "/" + svg_filename + "_" + "marked_shortest_stroke" + ".svg", "w")
doc.writexml(output)
output.close()
except Exception as e:
print(e)
if verbose == True:
print("Details are as follows.")
print("{} strokes have been marked.".format(marked_count))
print("The produced sketch is stored in {}".format(output_folder))
print(torchsketch())
return marked_count
