mirror of
https://github.com/deepfakes/faceswap
synced 2025-06-07 10:43:27 -04:00
242 lines
7.4 KiB
Python
242 lines
7.4 KiB
Python
"""
|
|
Contains some simple tests.
|
|
The purpose of this tests is to detect crashes and hangs
|
|
but NOT to guarantee the corectness of the operations.
|
|
For this we want another set of testcases using pytest.
|
|
|
|
Due to my lazy coding, DON'T USE PATHES WITH BLANKS !
|
|
"""
|
|
|
|
import sys
|
|
from subprocess import check_call, CalledProcessError
|
|
import urllib
|
|
from urllib.request import urlretrieve
|
|
import os
|
|
from os.path import join as pathjoin, expanduser
|
|
|
|
FAIL_COUNT = 0
|
|
TEST_COUNT = 0
|
|
_COLORS = {
|
|
"FAIL": "\033[1;31m",
|
|
"OK": "\033[1;32m",
|
|
"STATUS": "\033[1;37m",
|
|
"BOLD": "\033[1m",
|
|
"ENDC": "\033[0m"
|
|
}
|
|
|
|
|
|
def print_colored(text, color="OK", bold=False):
|
|
""" Print colored text
|
|
This might not work on windows,
|
|
although travis runs windows stuff in git bash, so it might ?
|
|
"""
|
|
color = _COLORS.get(color, color)
|
|
fmt = '' if not bold else _COLORS['BOLD']
|
|
print(f"{color}{fmt}{text}{_COLORS['ENDC']}")
|
|
|
|
|
|
def print_ok(text):
|
|
""" Print ok in colored text """
|
|
print_colored(text, "OK", True)
|
|
|
|
|
|
def print_fail(text):
|
|
""" Print fail in colored text """
|
|
print_colored(text, "FAIL", True)
|
|
|
|
|
|
def print_status(text):
|
|
""" Print status in colored text """
|
|
print_colored(text, "STATUS", True)
|
|
|
|
|
|
def run_test(name, cmd):
|
|
""" run a test """
|
|
global FAIL_COUNT, TEST_COUNT # pylint:disable=global-statement
|
|
print_status(f"[?] running {name}")
|
|
print(f"Cmd: {' '.join(cmd)}")
|
|
TEST_COUNT += 1
|
|
try:
|
|
check_call(cmd)
|
|
print_ok("[+] Test success")
|
|
return True
|
|
except CalledProcessError as err:
|
|
print_fail(f"[-] Test failed with {err}")
|
|
FAIL_COUNT += 1
|
|
return False
|
|
|
|
|
|
def download_file(url, filename): # TODO: retry
|
|
""" Download a file from given url """
|
|
if os.path.isfile(filename):
|
|
print_status(f"[?] '{url}' already cached as '{filename}'")
|
|
return filename
|
|
try:
|
|
print_status(f"[?] Downloading '{url}' to '{filename}'")
|
|
video, _ = urlretrieve(url, filename)
|
|
return video
|
|
except urllib.error.URLError as err:
|
|
print_fail(f"[-] Failed downloading: {err}")
|
|
return None
|
|
|
|
|
|
def extract_args(detector, aligner, in_path, out_path, args=None):
|
|
""" Extraction command """
|
|
py_exe = sys.executable
|
|
_extract_args = (f"{py_exe} faceswap.py extract -i {in_path} -o {out_path} -D {detector} "
|
|
f"-A {aligner}")
|
|
if args:
|
|
_extract_args += f" {args}"
|
|
return _extract_args.split()
|
|
|
|
|
|
def train_args(model, model_path, faces, iterations=1, batchsize=2, extra_args=""):
|
|
""" Train command """
|
|
py_exe = sys.executable
|
|
args = (f"{py_exe} faceswap.py train -A {faces} -B {faces} -m {model_path} -t {model} "
|
|
f"-b {batchsize} -i {iterations} {extra_args}")
|
|
return args.split()
|
|
|
|
|
|
def convert_args(in_path, out_path, model_path, writer, args=None):
|
|
""" Convert command """
|
|
py_exe = sys.executable
|
|
conv_args = (f"{py_exe} faceswap.py convert -i {in_path} -o {out_path} -m {model_path} "
|
|
f"-w {writer}")
|
|
if args:
|
|
conv_args += f" {args}"
|
|
return conv_args.split() # Don't use pathes with spaces ;)
|
|
|
|
|
|
def sort_args(in_path, out_path, sortby="face", groupby="hist"):
|
|
""" Sort command """
|
|
py_exe = sys.executable
|
|
_sort_args = (f"{py_exe} tools.py sort -i {in_path} -o {out_path} -s {sortby} -g {groupby} -k")
|
|
return _sort_args.split()
|
|
|
|
|
|
def set_train_config(value):
|
|
""" Update the mixed_precision and autoclip values to given value
|
|
|
|
Parameters
|
|
----------
|
|
value: bool
|
|
The value to set the config parameters to.
|
|
"""
|
|
old_val, new_val = ("False", "True") if value else ("True", "False")
|
|
base_path = os.path.split(os.path.dirname(os.path.abspath(__file__)))[0]
|
|
train_ini = os.path.join(base_path, "config", "train.ini")
|
|
try:
|
|
cmd = ["sed", "-i", f"s/autoclip = {old_val}/autoclip = {new_val}/", train_ini]
|
|
check_call(cmd)
|
|
cmd = ["sed",
|
|
"-i",
|
|
f"s/mixed_precision = {old_val}/mixed_precision = {new_val}/",
|
|
train_ini]
|
|
check_call(cmd)
|
|
print_ok(f"Set autoclip and mixed_precision to `{new_val}`")
|
|
except CalledProcessError as err:
|
|
print_fail(f"[-] Test failed with {err}")
|
|
return False
|
|
|
|
|
|
def main():
|
|
""" Main testing script """
|
|
vid_src = "https://faceswap.dev/data/test.mp4"
|
|
img_src = "https://archive.org/download/GPN-2003-00070/GPN-2003-00070.jpg"
|
|
base_dir = pathjoin(expanduser("~"), "cache", "tests")
|
|
|
|
vid_base = pathjoin(base_dir, "vid")
|
|
img_base = pathjoin(base_dir, "imgs")
|
|
os.makedirs(vid_base, exist_ok=True)
|
|
os.makedirs(img_base, exist_ok=True)
|
|
py_exe = sys.executable
|
|
was_trained = False
|
|
|
|
vid_path = download_file(vid_src, pathjoin(vid_base, "test.mp4"))
|
|
if not vid_path:
|
|
print_fail("[-] Aborting")
|
|
sys.exit(1)
|
|
vid_extract = run_test(
|
|
"Extraction video with cv2-dnn detector and cv2-dnn aligner.",
|
|
extract_args("Cv2-Dnn", "Cv2-Dnn", vid_path, pathjoin(vid_base, "faces"))
|
|
)
|
|
|
|
img_path = download_file(img_src, pathjoin(img_base, "test_img.jpg"))
|
|
if not img_path:
|
|
print_fail("[-] Aborting")
|
|
sys.exit(1)
|
|
run_test(
|
|
"Extraction images with cv2-dnn detector and cv2-dnn aligner.",
|
|
extract_args("Cv2-Dnn", "Cv2-Dnn", img_base, pathjoin(img_base, "faces"))
|
|
)
|
|
|
|
if vid_extract:
|
|
run_test(
|
|
"Generate configs and test help output",
|
|
(
|
|
py_exe, "faceswap.py", "-h"
|
|
)
|
|
)
|
|
run_test(
|
|
"Sort faces.",
|
|
sort_args(
|
|
pathjoin(vid_base, "faces"), pathjoin(vid_base, "faces_sorted"),
|
|
sortby="face"
|
|
)
|
|
)
|
|
|
|
run_test(
|
|
"Rename sorted faces.",
|
|
(
|
|
py_exe, "tools.py", "alignments", "-j", "rename",
|
|
"-a", pathjoin(vid_base, "test_alignments.fsa"),
|
|
"-c", pathjoin(vid_base, "faces_sorted"),
|
|
)
|
|
)
|
|
set_train_config(True)
|
|
run_test(
|
|
"Train lightweight model for 1 iteration with WTL, AutoClip, MixedPrecion",
|
|
train_args("lightweight",
|
|
pathjoin(vid_base, "model"),
|
|
pathjoin(vid_base, "faces"),
|
|
iterations=1,
|
|
batchsize=1,
|
|
extra_args="-M"))
|
|
|
|
set_train_config(False)
|
|
was_trained = run_test(
|
|
"Train lightweight model for 1 iterations WITHOUT WTL, AutoClip, MixedPrecion",
|
|
train_args("lightweight",
|
|
pathjoin(vid_base, "model"),
|
|
pathjoin(vid_base, "faces"),
|
|
iterations=1,
|
|
batchsize=1))
|
|
|
|
if was_trained:
|
|
run_test(
|
|
"Convert video.",
|
|
convert_args(
|
|
vid_path, pathjoin(vid_base, "conv"),
|
|
pathjoin(vid_base, "model"), "ffmpeg"
|
|
)
|
|
)
|
|
|
|
run_test(
|
|
"Convert images.",
|
|
convert_args(
|
|
img_base, pathjoin(img_base, "conv"),
|
|
pathjoin(vid_base, "model"), "opencv"
|
|
)
|
|
)
|
|
|
|
if FAIL_COUNT == 0:
|
|
print_ok(f"[+] Failed {FAIL_COUNT}/{TEST_COUNT} tests.")
|
|
sys.exit(0)
|
|
else:
|
|
print_fail(f"[-] Failed {FAIL_COUNT}/{TEST_COUNT} tests.")
|
|
sys.exit(1)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
main()
|