object format of the program

This commit is contained in:
Q
2021-11-07 10:26:31 +02:00
parent 492708e8fe
commit ab6aafa42e
2 changed files with 360 additions and 280 deletions

View File

@@ -1,3 +1,3 @@
from tsmark.video_annotator import main from tsmark.video_annotator import main
VERSION = "0.1" VERSION = "0.2"

View File

@@ -5,142 +5,171 @@ import time
import argparse import argparse
def draw_time(frame, nr, fps, paused): class Marker:
def __init__(self):
self.get_options()
if not os.path.exists(self.opts.video):
raise FileNotFoundError("Video file missing!")
self.paused = False
self.read_next = False
self.show_info = True
self.auto_step = True
self.font = cv2.FONT_HERSHEY_SIMPLEX
self.frame_visu = []
self.max_res = (1280, 720)
try:
self.open()
self.calculate_res()
self.parse_timestamps()
self.loop()
except Exception as e:
exc_type, exc_obj, exc_tb = sys.exc_info()
fname = os.path.split(exc_tb.tb_frame.f_code.co_filename)[1]
print(exc_type, fname, exc_tb.tb_lineno)
raise e
def open(self):
self.video_reader = cv2.VideoCapture(self.opts.video)
self.frames = int(self.video_reader.get(cv2.CAP_PROP_FRAME_COUNT))
self.fps = self.video_reader.get(cv2.CAP_PROP_FPS)
self.spf = 1 / self.fps
self.video_length = self.frames * self.fps
def calculate_res(self):
self.video_res = [
int(self.video_reader.get(cv2.CAP_PROP_FRAME_WIDTH)),
int(self.video_reader.get(cv2.CAP_PROP_FRAME_HEIGHT)),
]
video_aspect = self.video_res[0] / self.video_res[1]
if self.video_res[0] > self.max_res[0]:
self.video_res[0] = int(self.max_res[0])
self.video_res[1] = int(self.video_res[0] / video_aspect)
if self.video_res[1] > self.max_res[1]:
self.video_res[1] = int(self.max_res[1])
self.video_res[0] = int(self.video_res[1] * video_aspect)
self.video_res = tuple(self.video_res)
self.bar_start = int(self.video_res[0] * 0.05)
self.bar_end = int(self.video_res[0] * 0.95)
self.bar_top = int(self.video_res[1] * 0.90)
self.bar_bottom = int(self.video_res[1] * 0.95)
def calculate_step(self):
now = time.time()
self.last_move = [x for x in self.last_move if x[1] > now - 3]
if len(self.last_move) == 0:
self.step = 1
self.last_move = []
return
lefts = sum([1 for x in self.last_move if x[0] == "l"])
rights = sum([1 for x in self.last_move if x[0] == "r"])
if lefts > 0 and rights > 0:
self.step = 1
self.last_move = []
return
count = max(lefts, rights)
if count < 5:
self.step = 1
else:
# x2 poly from 5:5 -> 15:180
self.step = 45 - 16.5 * count + 1.7 * count * count
self.step = min(self.step, 0.1 * self.video_length)
self.step = int(self.step)
def draw_bar(self, frame):
position = self.nr / self.frames
bar_position = int(self.bar_start + position * (self.bar_end - self.bar_start))
cv2.rectangle(
frame,
(self.bar_start, self.bar_top),
(self.bar_end, self.bar_bottom),
(255, 255, 255),
2,
)
for ts in self.stamps:
ts_pos = int(
self.bar_start + ts / self.frames * (self.bar_end - self.bar_start)
)
cv2.line(
frame,
(ts_pos, self.bar_top),
(ts_pos, self.bar_bottom),
(32, 32, 32),
3,
)
cv2.line(
frame,
(ts_pos, self.bar_top),
(ts_pos, self.bar_bottom),
(84, 255, 63),
1,
)
# cv2.line(frame, (bar_position, top), (bar_position, bottom), (32,32,32), 3)
cv2.line(
frame,
(bar_position, self.bar_top),
(bar_position, self.bar_bottom),
(63, 84, 255),
1,
)
self.shadow_text(
frame,
"1",
(self.bar_start - 7, self.bar_bottom + 20),
0.7,
2,
(255, 255, 255),
)
end_frame = self.format_time(self.frames - 1)
(text_width, text_height) = cv2.getTextSize(
end_frame, cv2.FONT_HERSHEY_SIMPLEX, 0.7, 2
)[0]
self.shadow_text(
frame,
end_frame,
(self.bar_end - text_width, self.bar_bottom + 20),
0.7,
2,
(255, 255, 255),
)
def draw_label(self, frame):
if not self.nr in self.stamps:
return
text = "{} #{}".format(self.nr + 1, self.stamps.index(self.nr) + 1)
bottom = 60
left = 10
self.shadow_text(frame, text, (left, bottom), 1, 2, (63, 84, 255))
def draw_time(self, frame):
left = 10 left = 10
bottom = 30 bottom = 30
formatted = "{} {}".format( formatted = "{} {}".format(
format_time(nr, fps), self.format_time(self.nr),
"||" if paused else "", "||" if self.paused else "",
) )
shadow_text(frame, formatted, (left, bottom), 1.1, 2, (255, 255, 255)) self.shadow_text(frame, formatted, (left, bottom), 1.1, 2, (255, 255, 255))
def format_time(self, nframe):
def shadow_text(frame, text, pos, size, thicc, color): seconds = int(nframe / self.fps)
font = cv2.FONT_HERSHEY_SIMPLEX frame = nframe % self.fps
cv2.putText( parts = int(100 * (frame / self.fps))
frame,
text,
pos,
font,
size,
(0, 0, 0),
2 * thicc,
cv2.LINE_AA,
)
cv2.putText(
frame,
text,
pos,
font,
size,
color,
thicc,
cv2.LINE_AA,
)
def draw_bar(frame, nr, frames, fps, stamps):
position = nr / frames
bar_start = int(frame.shape[1] * 0.05)
bar_end = int(frame.shape[1] * 0.95)
bar_position = int(bar_start + position * (bar_end - bar_start))
top = int(frame.shape[0] * 0.90)
bottom = int(frame.shape[0] * 0.95)
cv2.rectangle(frame, (bar_start, top), (bar_end, bottom), (255, 255, 255), 2)
for ts in stamps:
ts_pos = int(bar_start + ts / frames * (bar_end - bar_start))
cv2.line(frame, (ts_pos, top), (ts_pos, bottom), (32, 32, 32), 3)
cv2.line(frame, (ts_pos, top), (ts_pos, bottom), (84, 255, 63), 1)
# cv2.line(frame, (bar_position, top), (bar_position, bottom), (32,32,32), 3)
cv2.line(frame, (bar_position, top), (bar_position, bottom), (63, 84, 255), 1)
shadow_text(frame, "1", (bar_start - 7, bottom + 20), 0.7, 2, (255, 255, 255))
end_frame = format_time(frames - 1, fps)
(text_width, text_height) = cv2.getTextSize(
end_frame, cv2.FONT_HERSHEY_SIMPLEX, 0.7, 2
)[0]
shadow_text(
frame, end_frame, (bar_end - text_width, bottom + 20), 0.7, 2, (255, 255, 255)
)
def draw_label(frame, nr, stamps):
if not nr in stamps:
return
text = "{} #{}".format(nr + 1, stamps.index(nr) + 1)
bottom = 60
left = 10
shadow_text(frame, text, (left, bottom), 1, 2, (63, 84, 255))
def calculate_step(step, last_move, video_length):
now = time.time()
last_move = [x for x in last_move if x[1] > now - 3]
if len(last_move) == 0:
return 1, []
lefts = sum([1 for x in last_move if x[0] == "l"])
rights = sum([1 for x in last_move if x[0] == "r"])
if lefts > 0 and rights > 0:
return 1, []
count = max(lefts, rights)
if count < 5:
step = 1
else:
# x2 poly from 5:5 -> 15:180
step = 45 - 16.5 * count + 1.7 * count * count
step = min(step, 0.1 * video_length)
return int(step), last_move
def format_time(nframe, fps):
seconds = int(nframe / fps)
frame = nframe % fps
parts = int(100 * (frame / fps))
return time.strftime("%H:%M:%S", time.gmtime(seconds)) + ".%02d" % (parts) return time.strftime("%H:%M:%S", time.gmtime(seconds)) + ".%02d" % (parts)
def get_options(self):
def parse_time(timestr, fps):
"""return frames"""
colon_count = len(timestr.split(":")) - 1
if colon_count == 0:
secs = float(timestr)
return int(secs * fps)
if colon_count == 1:
mins, secstr = timestr.split(":", 1)
sec = float(secstr)
return int(fps * (int(mins) * 60 + sec))
if colon_count == 2:
hours, mins, secstr = timestr.split(":", 2)
sec = float(secstr)
return int(fps * (int(hours) * 3600 + int(mins) * 60 + sec))
raise ValueError("Cannot parse time definition {}".format(timestr))
def print_timestamps(stamps, fps, filename):
stamps.sort()
print("# Timestamps:")
for i, ts in enumerate(stamps):
print("# {}: {} / {}".format(i + 1, format_time(ts, fps), ts + 1))
if len(stamps) > 0:
print(
"ffmpeg -i '{}' -ss {} -to {} -c copy trimmed.mp4".format(
filename,
format_time(stamps[0], fps),
format_time(stamps[-1], fps),
)
)
def get_options():
parser = argparse.ArgumentParser(description="Video timestamping tool") parser = argparse.ArgumentParser(description="Video timestamping tool")
parser.add_argument( parser.add_argument(
@@ -152,175 +181,226 @@ def get_options():
help="Comma separated list of predefined timestamps, in frame numbers, or HH:MM:SS.FF", help="Comma separated list of predefined timestamps, in frame numbers, or HH:MM:SS.FF",
) )
parser.add_argument(action="store", dest="video") parser.add_argument(action="store", dest="video")
return parser.parse_args() self.opts = parser.parse_args()
def mouse_click(self, event, x, y, flags, param):
def calculate_res(max_res, video_reader): in_bar = all(
(
x < self.bar_end,
x > self.bar_start,
y < self.bar_bottom,
y > self.bar_top,
)
)
if event == cv2.EVENT_LBUTTONDOWN:
if in_bar:
click_relative = (x - self.bar_start) / (self.bar_end - self.bar_start)
self.nr = int(click_relative * self.frames)
self.read_next = True
video_res = [ if event == cv2.EVENT_LBUTTONDBLCLK:
int(video_reader.get(cv2.CAP_PROP_FRAME_WIDTH)), if not in_bar:
int(video_reader.get(cv2.CAP_PROP_FRAME_HEIGHT)), self.toggle_stamp()
] # doubleclick (toggle?)
video_aspect = video_res[0] / video_res[1] # ~ print("double", x, y)
if video_res[0] > max_res[0]:
video_res[0] = int(max_res[0])
video_res[1] = int(video_res[0] / video_aspect)
if video_res[1] > max_res[1]: def parse_time(self, timestr):
video_res[1] = int(max_res[1]) """return frames"""
video_res[0] = int(video_res[1] * video_aspect)
return tuple(video_res)
colon_count = len(timestr.split(":")) - 1
if colon_count == 0:
secs = float(timestr)
return int(secs * self.fps)
if colon_count == 1:
mins, secstr = timestr.split(":", 1)
sec = float(secstr)
return int(self.fps * (int(mins) * 60 + sec))
if colon_count == 2:
hours, mins, secstr = timestr.split(":", 2)
sec = float(secstr)
return int(self.fps * (int(hours) * 3600 + int(mins) * 60 + sec))
raise ValueError("Cannot parse time definition {}".format(timestr))
def print_help(): def parse_timestamps(self):
if self.opts.timestamps:
self.stamps = sorted(
[self.parse_time(ts.strip()) for ts in self.opts.timestamps.split(",")]
)
self.stamps = [x for x in self.stamps if 0 <= x < self.frames]
self.nr = self.stamps[0]
else:
self.stamps = []
self.nr = 0
def print_help(self):
print( print(
"""Keyboard help: """Keyboard help:
(Note: after mouse click, arrows stop working due to unknown bug: use j,k)
Arrows left and right, Home, End Arrows left and right, Home, End or click mouse in position bar
j and k
jump in video. Tap frequently to increase time step jump in video. Tap frequently to increase time step
, and . move one frame at a time , and . move one frame at a time
z and c move to previous or next mark z and c move to previous or next mark
x mark frame x or double click in the video
mark frame
space pause space pause
i toggle HUD i toggle HUD
q quit q quit
""" """
) )
def print_timestamps(self):
def main(): self.stamps.sort()
try: print("# Timestamps:")
opts = get_options() for i, ts in enumerate(self.stamps):
if not os.path.exists(opts.video): print("# {}: {} / {}".format(i + 1, self.format_time(ts), ts + 1))
raise FileNotFoundError("Video file missing!") if len(self.stamps) > 0:
print(
max_res = (1280, 720) "ffmpeg -i '{}' -ss {} -to {} -c copy trimmed.mp4".format(
video_reader = cv2.VideoCapture(opts.video) self.opts.video,
frames = int(video_reader.get(cv2.CAP_PROP_FRAME_COUNT)) self.format_time(self.stamps[0]),
fps = video_reader.get(cv2.CAP_PROP_FPS) self.format_time(self.stamps[-1]),
spf = 1.0 / fps
video_res = calculate_res(max_res, video_reader)
if opts.timestamps:
stamps = sorted(
[parse_time(ts.strip(), fps) for ts in opts.timestamps.split(",")]
) )
stamps = [x for x in stamps if 0 <= x < frames] )
nr = stamps[0]
def shadow_text(self, frame, text, pos, size, thicc, color):
cv2.putText(
frame,
text,
pos,
self.font,
size,
(0, 0, 0),
2 * thicc,
cv2.LINE_AA,
)
cv2.putText(
frame,
text,
pos,
self.font,
size,
color,
thicc,
cv2.LINE_AA,
)
def toggle_stamp(self):
if self.nr in self.stamps:
self.stamps.remove(self.nr)
else: else:
stamps = [] self.stamps.append(self.nr)
nr = 0 self.stamps.sort()
paused = False def loop(self):
read_next = False
show_info = True
frame_visu = []
step = 1
last_move = []
video_length = frames * fps
auto_step = True
video_reader.set(cv2.CAP_PROP_POS_FRAMES, nr)
print_help() self.step = 1
self.last_move = []
self.video_reader.set(cv2.CAP_PROP_POS_FRAMES, self.nr)
while video_reader.isOpened(): self.print_help()
cv2.namedWindow("tsmark")
cv2.setMouseCallback("tsmark", self.mouse_click)
while self.video_reader.isOpened():
show_time = time.time() show_time = time.time()
if (not paused) or read_next: if (not self.paused) or self.read_next:
ret, frame = video_reader.read() ret, frame = self.video_reader.read()
if ret == True: if ret == True:
if paused: if self.paused:
draw_wait = 200 draw_wait = 200
else: else:
draw_wait = 1 draw_wait = 1
if (not paused) or read_next: if (not self.paused) or self.read_next:
read_next = False self.read_next = False
frame_visu = cv2.resize(frame.copy(), video_res) frame_visu = cv2.resize(frame.copy(), self.video_res)
nr_time = nr / fps nr_time = self.nr / self.fps
if show_info: if self.show_info:
draw_time(frame_visu, nr, fps, paused) self.draw_time(frame_visu)
draw_bar(frame_visu, nr, frames, fps, stamps) self.draw_bar(frame_visu)
draw_label(frame_visu, nr, stamps) self.draw_label(frame_visu)
cv2.imshow("tsmark", frame_visu) cv2.imshow("tsmark", frame_visu)
k = cv2.waitKey(draw_wait) k = cv2.waitKey(draw_wait)
if k & 0xFF == ord("q") or k & 0xFF == 27: if k & 0xFF == ord("q") or k & 0xFF == 27:
break break
if k & 0xFF == 32: # space if k & 0xFF == 32: # space
paused = not paused self.paused = not self.paused
if k & 0xFF == 80: # home key if k & 0xFF == 80: # home key
nr = 0 self.nr = -1
self.read_next = True
if k & 0xFF == 87: # end key if k & 0xFF == 87: # end key
nr = frames - 1 self.nr = self.frames - 1
paused = True self.paused = True
self.read_next = True
if k & 0xFF == 83: # right arrow if k & 0xFF == 83 or k & 0xFF == ord("k"): # right arrow
last_move.append(("r", time.time())) self.last_move.append(("r", time.time()))
if auto_step: if self.auto_step:
step, last_move = calculate_step(step, last_move, video_length) self.calculate_step()
nr = int((nr_time + step) * fps) - 1 self.nr = int((nr_time + self.step) * self.fps) - 1
read_next = True self.read_next = True
if k & 0xFF == ord("."): if k & 0xFF == ord("."):
paused = True self.paused = True
read_next = True self.read_next = True
if k & 0xFF == 81: # left arrow if k & 0xFF == 81 or k & 0xFF == ord("j"): # left arrow
last_move.append(("l", time.time())) self.last_move.append(("l", time.time()))
if auto_step: if self.auto_step:
step, last_move = calculate_step(step, last_move, video_length) self.calculate_step()
nr = int((nr_time - step) * fps) - 1 self.nr = int((nr_time - self.step) * self.fps) - 1
read_next = True self.read_next = True
if k & 0xFF == ord(","): if k & 0xFF == ord(","):
paused = True self.paused = True
nr -= 2 self.nr -= 2
read_next = True self.read_next = True
if k & 0xFF == ord("z"): # move to previous ts if k & 0xFF == ord("z"): # move to previous ts
for ts in reversed(sorted(stamps)): for ts in reversed(sorted(self.stamps)):
if ts < nr - 1: if ts < self.nr - 1:
nr = ts - 1 self.nr = ts - 1
read_next = True self.read_next = True
break break
if k & 0xFF == ord("c"): # move to previous ts if k & 0xFF == ord("c"): # move to previous ts
for ts in sorted(stamps): for ts in sorted(self.stamps):
if ts > nr: if ts > self.nr:
nr = ts - 1 self.nr = ts - 1
read_next = True self.read_next = True
break break
if k & 0xFF == ord("x"): # toggle ts if k & 0xFF == ord("x"): # toggle ts
if nr in stamps: self.toggle_stamp()
stamps.remove(nr)
else:
stamps.append(nr)
stamps.sort()
if k & 0xFF == ord("i"): if k & 0xFF == ord("i"):
show_info = not show_info self.show_info = not self.show_info
if k & 0xFF == ord("h"): if k & 0xFF == ord("h"):
print_help() self.print_help()
if (not paused) or read_next: if (not self.paused) or self.read_next:
nr += 1 self.nr += 1
if nr < 0: if self.nr < 0:
nr = 0 self.nr = 0
if nr >= frames: if self.nr >= self.frames:
nr = frames - 1 self.nr = self.frames - 1
paused = True self.paused = True
if read_next: if self.read_next:
video_reader.set(cv2.CAP_PROP_POS_FRAMES, nr) self.video_reader.set(cv2.CAP_PROP_POS_FRAMES, self.nr)
time_to_wait = spf - time.time() + show_time time_to_wait = self.spf - time.time() + show_time
if time_to_wait > 0: if time_to_wait > 0:
time.sleep(time_to_wait) time.sleep(time_to_wait)
else: else:
nr = frames - 2 self.nr = self.frames - 2
video_reader.set(cv2.CAP_PROP_POS_FRAMES, nr) self.video_reader.set(cv2.CAP_PROP_POS_FRAMES, self.nr)
paused = True self.paused = True
read_next = True self.read_next = True
video_reader.release() self.video_reader.release()
print_timestamps(stamps, fps, opts.video) self.print_timestamps()
except Exception:
exc_type, exc_obj, exc_tb = sys.exc_info()
fname = os.path.split(exc_tb.tb_frame.f_code.co_filename)[1] def main():
print(exc_type, fname, exc_tb.tb_lineno) mark = Marker()