def main():
from overtrack_cv.util.logging_config import config_logger
from overtrack_cv.util.test_processor import test_processor
config_logger(os.path.basename(__file__),
level=logging.DEBUG,
write_to_file=False)
proc = TopHudProcessor()
agent_paths = [
os.path.join(
"C:/Users/simon/overtrack_2/valorant_images/top_hud_agents",
agent_name.lower() + ".png",
) for agent_name in agents
]
agent_frames = [
agent_path for agent_path in agent_paths if os.path.exists(agent_path)
]
# util.test_processor(agent_frames, proc, 'valorant.top_hud', test_all=False, wait=True)
# util.test_processor([
# r'C:/Users/simon/overtrack_2/valorant_images/ingame\00-41-617.image.png'
# ], TopHudProcessor(), 'valorant.top_hud', test_all=False, wait=True)
#
test_processor(proc, "valorant.top_hud", test_all=False, wait=True)
def main():
from overtrack_cv.util.logging_config import config_logger
from overtrack_cv.util.test_processor import test_processor
config_logger(os.path.basename(__file__), level=logging.DEBUG, write_to_file=False)
test_processor(HomeScreenProcessor(), "valorant.home_screen")
def main():
from overtrack_cv.util.logging_config import config_logger
from overtrack_cv.util.test_processor import test_processor
config_logger(os.path.basename(__file__),
level=logging.DEBUG,
write_to_file=False)
test_processor(AgentSelectProcessor(),
"valorant.agent_select",
test_all=False)
def main():
from overtrack_cv.util.logging_config import config_logger
from overtrack_cv.util.test_processor import test_processor
config_logger(os.path.basename(__file__),
level=logging.DEBUG,
write_to_file=False)
proc = KillfeedProcessor()
test_processor(proc, "valorant.killfeed", wait=True, test_all=False)
test_processor(proc, "valorant.killfeed", wait=True, test_all=False)
def main():
from overtrack_cv.util.logging_config import config_logger
from overtrack_cv.util.test_processor import test_processor
config_logger(os.path.basename(__file__),
level=logging.DEBUG,
write_to_file=False)
p = PostgameProcessor()
test_processor(p,
"valorant.postgame",
"valorant.scoreboard",
test_all=False)
test_processor(p,
"valorant.postgame",
"valorant.scoreboard",
test_all=False)
def main() -> None:
config_logger("shmem_source", logging.DEBUG, False)
cap = SharedMemoryCapture(debug_frames=True)
cap.start()
while True:
frame = cap.get()
if not frame:
print("x")
time.sleep(4)
continue
elif isinstance(frame, Exception):
raise frame
cv2.imshow("frame", frame.debug_image)
print(frame.source)
cv2.waitKey(1000)
def main() -> None:
convert_tflite_model()
config_logger("big_noodle", logging.DEBUG, False)
# from tensorflow.python.keras import Input
# from tensorflow.python.keras.layers import Reshape, Conv2D
# from tensorflow.python.keras import Model
# from tensorflow.python.keras.layers import Activation
# from overtrack.util.custom_layers import MaxAlongDims
# from tensorflow.python.keras.layers import Layer
# from tensorflow.python.keras.saving import export_saved_model
#
# image = Input(shape=(Classifier.HEIGHT, Classifier.WIDTH), name='images', dtype=tf.float32)
# image_r = Reshape((Classifier.HEIGHT, Classifier.WIDTH, 1))(image)
# conv1 = Conv2D(
# len(Classifier.CHARACTERS),
# Classifier.CONV1_SIZE,
# activation=None,
# name='conv1'
# )(image_r)
# logits = MaxAlongDims((1, 2))(conv1)
# output = Activation(activation='softmax')(logits)
#
# model = Model(
# inputs=image,
# outputs=output
# )
# model.compile(
# optimizer='adam',
# loss='binary_crossentropy'
# )
#
# model.summary()
#
# l: Layer = model.layers[2]
#
# p = os.path.join(os.path.dirname(__file__), 'data', 'big_noodle.npz')
# restore_dict = dict(np.load(p))
# l.set_weights([restore_dict['conv1/kernel:0'], restore_dict['conv1/bias:0']])
#
# export_saved_model(model, f'./data/big_noodle', serving_only=True)
image = cv2.imread("C:/tmp/test.png")
for _ in range(10):
print(ocr(image))
def main() -> None:
convert_tflite_model()
config_logger("digit", logging.DEBUG, False)
# from tensorflow.python.keras import Input
# from tensorflow.python.keras.layers import Reshape
# from tensorflow.python.keras import Model
# from tensorflow.python.keras.layers import Layer
# from tensorflow.python.keras.layers import Dense
# from tensorflow.python.keras.saving import export_saved_model
#
# image = Input(shape=(Classifier.IMSIZE[0], Classifier.IMSIZE[1]), name='images', dtype=tf.float32)
# image_r = Reshape((np.prod(Classifier.IMSIZE), ))(image)
#
# logits = Dense(
# len(Classifier.CHARACTERS),
# activation='softmax',
# name='fc1'
# )(image_r)
#
# model = Model(
# inputs=image,
# outputs=logits
# )
# model.compile(
# optimizer='adam',
# loss='binary_crossentropy'
# )
#
# model.summary()
#
# p = os.path.join(os.path.dirname(__file__), 'data', 'tab_ocr.npz')
# restore_dict = dict(np.load(p))
# print({k: v.shape for k, v in restore_dict.items()})
# l: Layer = model.layers[2]
# print(l, [w.shape for w in l.get_weights()])
# l.set_weights([restore_dict['tab_ocr/W_fc_:0'], restore_dict['tab_ocr/b_fc_:0']])
#
# os.makedirs('./data/tab', exist_ok=True)
# export_saved_model(model, f'./data/tab_digit', serving_only=True)
image = cv2.imread("C:/tmp/test.png")
for _ in range(10):
print(ocr(image))
def main() -> None:
import glob
config_logger("map_processor", logging.DEBUG, write_to_file=False)
# find_best_match()
# exit(0)
pipeline = MinimapProcessor()
paths = glob.glob("d:/overtrack/frames_rot/*.json")[1630:]
# random.shuffle(paths)
# paths = ['D:/overtrack/frames_rot/1569668416.76_image.png']
# paths = [
# r'd:/overtrack/frames_rot\1569669144.44_frame.json'
# r'd:/overtrack/frames_rot\1569670312.41_image.png'
# ]
for p in paths:
print(p)
imp = p.replace("frame.json", "image.png")
try:
frame = Frame.create(
cv2.resize(cv2.imread(imp), (1920, 1080)),
float(os.path.basename(p).split("_")[0]),
True,
file=imp,
)
except:
logging.exception("")
continue
frame.game_time = frame.timestamp - 1569665630
pipeline.process(frame)
print(frame)
cv2.imshow("debug", frame.debug_image)
cv2.waitKey(0)
def main():
from overtrack_cv.util.logging_config import config_logger
from overtrack_cv.util.test_processor import test_processor
config_logger(os.path.basename(__file__), level=logging.DEBUG, write_to_file=False)
test_processor(TimerProcessor(), "timer", wait=True)
elif errorcode in ["0x887A0004", "0x80070057"]:
return "Unsupported by Graphics Driver:\nYou may need to run OverTrack on the a different GPU, or your device may not be supported"
return ""
def main() -> None:
# hwnd, pid = get_hwnd('Overwatch', None)
# hwnd, pid = get_hwnd(None, 'pycharm64.exe')
# print(hwnd, pid)
from overtrack_models.dataclasses.typedload import referenced_typedload
# d = DisplayDuplicationCapture(window_title='Overwatch', executable='overwatch.exe')
title = "Apex Legends"
executable = "r5apex.exe"
d = DisplayDuplicationCapture(window_title=title, executable=executable)
while True:
frame = d.get()
cv2.imshow("frame", frame.image)
frame.strip()
print(frame.valid)
cv2.waitKey(1000)
referenced_typedload.dump(frame)
if __name__ == "__main__":
config_logger("display_duplication", logging.DEBUG, False)
main()
def test_processor(
proc: Processor,
*fields: str,
show=True,
test_all=True,
wait=True,
warmup=True,
game=None,
images=None,
) -> None:
proc.eager_load()
if not game:
game = os.path.basename(
glob.glob(
os.path.join(os.path.dirname(inspect.getfile(proc.__class__)),
"..", "..", "*_frame_data.py"))[0]).split("_",
1)[0]
print(game)
print(inspect.getmodule(proc))
print(proc.__class__.__module__)
# TODO
game_data_factory = object
if game == "apex":
from overtrack_cv.games.apex.apex_frame_data import ApexFrameData
game_data_factory = ApexFrameData
elif game == "valorant":
from overtrack_cv_private.games.valorant.valorant_frame_data import (
ValorantFrameData, )
game_data_factory = ValorantFrameData
elif game == "overwatch":
from overtrack_cv_private.games.overwatch.overwatch_frame_data import (
OverwatchFrameData, )
game_data_factory = OverwatchFrameData
import numpy as np
if warmup:
for _ in range(10):
proc.run(
Frame.create(np.zeros((1080, 1920, 3), dtype=np.uint8),
0,
warmup=True,
**{game: game_data_factory()}))
if not images:
images = os.path.join(os.path.dirname(inspect.getfile(proc.__class__)),
"samples", "*")
if isinstance(images, list):
config_logger("test_processor", logging.DEBUG, False)
paths = images
else:
config_logger(os.path.basename(images), logging.DEBUG, False)
if "*" in images:
paths = glob.glob(images)
elif images.endswith(".png"):
paths = [images]
elif images[1] == ":":
paths = glob.glob(images + "/*")
else:
paths = glob.glob(os.path.join(images, "*"))
paths.sort(key=lambda p: os.path.getmtime(p), reverse=True)
if test_all:
paths += glob.glob(
os.path.join(
os.path.dirname(__file__),
"..",
"games",
game or ".",
"**",
"samples",
"*.png",
))
for p in paths:
if not (p.endswith(".png") or p.endswith(".jpg")
or p.endswith(".jpeg")):
continue
time.sleep(0.01)
print("\n\n" + "-" * 32)
print(os.path.abspath(p))
im = cv2.imread(p)
ratio = im.shape[0] / im.shape[1]
if round(ratio, 2) != 0.56:
logger.warning(
f"Ignoring {p} had invalid aspect ratio - dimensions were {im.shape}"
)
continue
im = cv2.resize(im, (1920, 1080))
f = Frame.create(im,
os.path.getctime(p),
debug=True,
**{game: game_data_factory()})
f.source_image = p
if "game_time=" in p:
f.game_time = float(
os.path.basename(p).split("=", 1)[1].rsplit(".", 1)[0])
f.current_game = CurrentGame()
stats = proc.run(f)
debug_image = f.debug_image
f.strip()
# print(json.dumps(frameload.frames_dump(f, numpy_support=True), indent=2))
def getval(f: Frame, n: str):
v = f
for p in n.split("."):
v = getattr(v, p, None)
if not v:
return None
return v
if game:
fields = list(fields) # + [game]
# def pformat(v):
print(
tabulate.tabulate(
[("source_image", p), ("processor_result", stats.result)] +
[(f"{game}.{fl}", getattr(getattr(f, game), fl, None))
for fl in fields]
# +
# [
# (
# n,
# pformat(getval(f, n))
# )
# for n in fields
# ]
))
pprint(stats)
print(stats.duration * 1000)
# if game:
# pprint(getattr(f, game))
# print()
if show:
cv2.imshow("debug", debug_image)
if wait == "sometimes":
c = cv2.waitKey(0 if stats.result else 1)
else:
c = cv2.waitKey(0 if wait else 1)
if c == 115:
os.remove(p)
print("-" * 32)
def main() -> None:
config_logger("big_noodle", logging.DEBUG, False)
# from tensorflow.python.keras import Input
# from tensorflow.python.keras.layers import Conv2D
# from tensorflow.python.keras import Model
# from tensorflow.python.keras.layers import Layer
# from tensorflow.python.keras.saving import export_saved_model
# from tensorflow.python.keras.layers import Lambda
#
# image = Input(shape=(Decoder.IMAGE_HEIGHT, Decoder.IMAGE_WIDTH, 3), name='images', dtype=tf.float32)
# image_norm = Lambda(lambda i: i / 255.0 - 0.5)(image)
# conv1 = Conv2D(
# Decoder.CONV1_LAYERS,
# Decoder.CONV1_SIZE,
# activation='relu',
# padding='same',
# name='conv1'
# )(image_norm)
# conv1_pad = Pad(
# ((0, 0), (0, 0), (0, Decoder.CONV2_WIDTH - 1), (0, 0))
# )(conv1)
# conv2 = Conv2D(
# len(Decoder.CHARACTERS) + 1,
# (Decoder.IMAGE_HEIGHT - 2, Decoder.CONV2_WIDTH),
# activation=None,
# padding='valid',
# name='conv2'
# )(conv1_pad)
# logits = MaxAlongDims(dims=(1, ))(conv2)
#
# model = Model(
# inputs=image,
# outputs=logits
# )
# model.compile(
# optimizer='adam',
# loss='binary_crossentropy'
# )
# model.summary()
#
# p = os.path.join(os.path.dirname(__file__), 'data', 'big_noodle_ctc.npz')
# restore_dict = dict(np.load(p))
# print({k: v.shape for k, v in restore_dict.items()})
#
# l: Layer = model.layers[2]
# print(l, [w.shape for w in l.get_weights()])
# l.set_weights([restore_dict['conv1/kernel:0'], restore_dict['conv1/bias:0']])
#
# l: Layer = model.layers[4]
# print(l, [w.shape for w in l.get_weights()])
# l.set_weights([restore_dict['conv2/kernel:0'], restore_dict['conv2/bias:0']])
#
# os.makedirs('./data/big_noodle_ctc')
# export_saved_model(model, f'./data/big_noodle_ctc', serving_only=True)
image = cv2.imread("C:/tmp/test.png")
for _ in range(10):
print(ocr(image))
for _ in range(10):
print(ocr_all([image, np.hstack((image, image))]))