Robert Cranston authored on 04/06/2020 17:59:34
| 1 | 1 |
new file mode 100644 |
| ... | ... |
@@ -0,0 +1,391 @@ |
| 1 |
+""" |
|
| 2 |
+Python library to ease processing of video frames. |
|
| 3 |
+""" |
|
| 4 |
+ |
|
| 5 |
+import sys |
|
| 6 |
+import os |
|
| 7 |
+import signal |
|
| 8 |
+import argparse |
|
| 9 |
+import json |
|
| 10 |
+import datetime |
|
| 11 |
+import shlex |
|
| 12 |
+import ffmpeg # ffmpeg-python |
|
| 13 |
+import numpy as np |
|
| 14 |
+import PIL.Image |
|
| 15 |
+import cv2 |
|
| 16 |
+ |
|
| 17 |
+ |
|
| 18 |
+def error(message): |
|
| 19 |
+ print(f"{sys.argv[0]}: Error: {message}", file=sys.stderr)
|
|
| 20 |
+ exit(1) |
|
| 21 |
+ |
|
| 22 |
+ |
|
| 23 |
+def resize(image, width, height, resample): |
|
| 24 |
+ return np.array( |
|
| 25 |
+ PIL.Image |
|
| 26 |
+ .fromarray(image) |
|
| 27 |
+ .resize((width, height), resample) |
|
| 28 |
+ ) |
|
| 29 |
+ |
|
| 30 |
+ |
|
| 31 |
+def text(image, text, x, y, thickness, font=cv2.FONT_HERSHEY_SIMPLEX): |
|
| 32 |
+ margin = 10 |
|
| 33 |
+ height, width = image.shape[:2] |
|
| 34 |
+ size, _ = cv2.getTextSize(text, font, 1, 1) |
|
| 35 |
+ scale = min(1, (width - 2 * margin) / size[0]) |
|
| 36 |
+ x = np.clip(x, margin, width - int(scale * size[0]) - margin) |
|
| 37 |
+ y = np.clip(y, margin + int(scale * size[1]), height - margin) |
|
| 38 |
+ black = (0, 0, 0) |
|
| 39 |
+ white = (255, 255, 255) |
|
| 40 |
+ image = cv2.putText(image, text, (x, y), font, scale, black, 3 * thickness) |
|
| 41 |
+ image = cv2.putText(image, text, (x, y), font, scale, white, thickness) |
|
| 42 |
+ return image |
|
| 43 |
+ |
|
| 44 |
+ |
|
| 45 |
+def fix_rect(args, x, y, w, h): |
|
| 46 |
+ # Grow into correct aspect ratio. |
|
| 47 |
+ nw = max(w, int(h * args.output_aspect)) |
|
| 48 |
+ nh = max(h, int(w / args.output_aspect)) |
|
| 49 |
+ # Optionally shrink into working image size. |
|
| 50 |
+ scale = min(1, args.working_width / nw, args.working_height / nh) |
|
| 51 |
+ nw = int(nw * scale) |
|
| 52 |
+ nh = int(nh * scale) |
|
| 53 |
+ # Center new region on old region. |
|
| 54 |
+ nx = x - (nw - w) // 2 |
|
| 55 |
+ ny = y - (nh - h) // 2 |
|
| 56 |
+ # Push into working image. |
|
| 57 |
+ nx = max(0, nx) - max(0, (nx+nw) - args.working_width) |
|
| 58 |
+ ny = max(0, ny) - max(0, (ny+nh) - args.working_height) |
|
| 59 |
+ # Return. |
|
| 60 |
+ return nx, ny, nw, nh |
|
| 61 |
+ |
|
| 62 |
+ |
|
| 63 |
+def run(description, process, init=None, args_pre=None, args_post=None): |
|
| 64 |
+ # Parse arguments |
|
| 65 |
+ parser = argparse.ArgumentParser(description=description) |
|
| 66 |
+ if args_pre is not None: |
|
| 67 |
+ args_pre(parser) |
|
| 68 |
+ parser.add_argument( |
|
| 69 |
+ '--start-frame', metavar='FRAME', type=int, |
|
| 70 |
+ help="starting input frame to process (inclusive)") |
|
| 71 |
+ parser.add_argument( |
|
| 72 |
+ '--end-frame', metavar='FRAME', type=int, |
|
| 73 |
+ help="ending input frame to process (exclusive)") |
|
| 74 |
+ parser.add_argument( |
|
| 75 |
+ '--decimate', metavar="COUNT", type=int, default=1, |
|
| 76 |
+ help="only use every %(metavar)s input frame") |
|
| 77 |
+ parser.add_argument( |
|
| 78 |
+ '--width', type=int, |
|
| 79 |
+ help="width of output video") |
|
| 80 |
+ parser.add_argument( |
|
| 81 |
+ '--height', type=int, |
|
| 82 |
+ help="height of output video") |
|
| 83 |
+ parser.add_argument( |
|
| 84 |
+ '--codec', |
|
| 85 |
+ help="codec of output video" |
|
| 86 |
+ ) |
|
| 87 |
+ parser.add_argument( |
|
| 88 |
+ '--pix-fmt', |
|
| 89 |
+ help="pixel format of output video") |
|
| 90 |
+ parser.add_argument( |
|
| 91 |
+ '--extra-args', metavar='ARGS', |
|
| 92 |
+ help=""" |
|
| 93 |
+ extra arguments to pass to FFmpeg, as a JSON object (do not specify the |
|
| 94 |
+ leading dash for keys, use a null value for arguments that do not take |
|
| 95 |
+ a parameter) |
|
| 96 |
+ """) |
|
| 97 |
+ parser.add_argument( |
|
| 98 |
+ '--no-preview', action='store_true', |
|
| 99 |
+ help="do not show any previews") |
|
| 100 |
+ parser.add_argument( |
|
| 101 |
+ '--no-write', action='store_true', |
|
| 102 |
+ help="do not write any files") |
|
| 103 |
+ parser.add_argument( |
|
| 104 |
+ '--no-audio', action='store_true', |
|
| 105 |
+ help="do not include audio in output files") |
|
| 106 |
+ parser.add_argument( |
|
| 107 |
+ '--overwrite', action='store_true', |
|
| 108 |
+ help="overwrite output files") |
|
| 109 |
+ parser.add_argument( |
|
| 110 |
+ '--debug', metavar='DEBUG_FILE', |
|
| 111 |
+ help="produce debug output (leave empty to base filename on input)") |
|
| 112 |
+ parser.add_argument( |
|
| 113 |
+ '--output', metavar='OUTPUT_FILE', |
|
| 114 |
+ help="produce final output (leave empty to base filename on input)") |
|
| 115 |
+ parser.add_argument( |
|
| 116 |
+ 'input', metavar='INPUT_FILE', |
|
| 117 |
+ help="input file") |
|
| 118 |
+ if args_post is not None: |
|
| 119 |
+ args_post(parser) |
|
| 120 |
+ args = parser.parse_args() |
|
| 121 |
+ |
|
| 122 |
+ # Check arguments. |
|
| 123 |
+ root, ext = os.path.splitext(os.path.basename(args.input)) |
|
| 124 |
+ for file in ["debug", "output"]: |
|
| 125 |
+ if getattr(args, file) == "": |
|
| 126 |
+ setattr(args, file, f"{root}_{file}{ext}")
|
|
| 127 |
+ if os.path.isfile(args.output) and not args.overwrite: |
|
| 128 |
+ error(f"File exists: '{args.output}', use --overwrite to overwrite.")
|
|
| 129 |
+ if os.path.isfile(args.debug) and not args.overwrite: |
|
| 130 |
+ error(f"File exists: '{args.debug}', use --overwrite to overwrite.")
|
|
| 131 |
+ if not os.path.isfile(args.input): |
|
| 132 |
+ error(f"File does not exist: '{args.input}'.")
|
|
| 133 |
+ |
|
| 134 |
+ # Probe input. |
|
| 135 |
+ probe = ffmpeg.probe(args.input, select_streams='v') |
|
| 136 |
+ |
|
| 137 |
+ # Common parameters. |
|
| 138 |
+ args.duration = float(probe['format']['duration']) |
|
| 139 |
+ args.frame_count = int(probe['streams'][0]['nb_frames']) |
|
| 140 |
+ |
|
| 141 |
+ # Input parameters. |
|
| 142 |
+ args.input_codec = probe['streams'][0]['codec_name'] |
|
| 143 |
+ args.input_pix_fmt = probe['streams'][0]['pix_fmt'] |
|
| 144 |
+ args.input_width = probe['streams'][0]['width'] |
|
| 145 |
+ args.input_height = probe['streams'][0]['height'] |
|
| 146 |
+ args.input_aspect = args.input_width / args.input_height |
|
| 147 |
+ args.input_fps = (lambda x, y: x / y)( |
|
| 148 |
+ *map(int, probe['streams'][0]['r_frame_rate'].split("/"))
|
|
| 149 |
+ ) |
|
| 150 |
+ |
|
| 151 |
+ # Output parameters. |
|
| 152 |
+ args.output_codec = args.codec or args.input_codec |
|
| 153 |
+ args.output_pix_fmt = args.pix_fmt or args.input_pix_fmt |
|
| 154 |
+ args.output_width = args.width or args.input_width |
|
| 155 |
+ args.output_height = args.height or int( |
|
| 156 |
+ args.output_width / args.input_aspect |
|
| 157 |
+ ) |
|
| 158 |
+ args.output_aspect = args.output_width / args.output_height |
|
| 159 |
+ args.output_fps = args.input_fps / (args.decimate or 1) |
|
| 160 |
+ |
|
| 161 |
+ # Working parameters. |
|
| 162 |
+ args.working_width = min( |
|
| 163 |
+ args.input_width, |
|
| 164 |
+ 2 * max( |
|
| 165 |
+ args.output_width, |
|
| 166 |
+ int(args.output_height * args.input_aspect) |
|
| 167 |
+ ) |
|
| 168 |
+ ) |
|
| 169 |
+ args.working_height = min( |
|
| 170 |
+ args.input_height, |
|
| 171 |
+ 2 * max( |
|
| 172 |
+ args.output_height, |
|
| 173 |
+ int(args.output_width / args.input_aspect) |
|
| 174 |
+ ) |
|
| 175 |
+ ) |
|
| 176 |
+ args.working_width += args.working_width % 2 |
|
| 177 |
+ args.working_height += args.working_height % 2 |
|
| 178 |
+ args.thickness = max(1, int(args.working_width / 1000)) |
|
| 179 |
+ |
|
| 180 |
+ # Fill in default arguments. |
|
| 181 |
+ if args.start_frame is None: |
|
| 182 |
+ args.start_frame = 0 |
|
| 183 |
+ if args.end_frame is None: |
|
| 184 |
+ args.end_frame = args.frame_count |
|
| 185 |
+ if args.extra_args is None: |
|
| 186 |
+ args.extra_args = {}
|
|
| 187 |
+ else: |
|
| 188 |
+ try: |
|
| 189 |
+ args.extra_args = json.loads(args.extra_args) |
|
| 190 |
+ except json.decoder.JSONDecodeError: |
|
| 191 |
+ error(f"Extra arguments is not valid JSON.") |
|
| 192 |
+ if type(args.extra_args) is not dict: |
|
| 193 |
+ error(f"Extra arguments is not a JSON object.") |
|
| 194 |
+ |
|
| 195 |
+ # Open files. |
|
| 196 |
+ debug_size = f'{args.working_width}x{args.working_height}'
|
|
| 197 |
+ output_size = f'{args.output_width}x{args.output_height}'
|
|
| 198 |
+ pipe_args = {
|
|
| 199 |
+ 'format': 'rawvideo', |
|
| 200 |
+ 'pix_fmt': 'rgb24', |
|
| 201 |
+ } |
|
| 202 |
+ output_args = {
|
|
| 203 |
+ 'vcodec': args.output_codec, |
|
| 204 |
+ 'pix_fmt': args.output_pix_fmt, |
|
| 205 |
+ 'shortest': None, |
|
| 206 |
+ **args.extra_args, |
|
| 207 |
+ } |
|
| 208 |
+ audio_args = ( |
|
| 209 |
+ [ |
|
| 210 |
+ ffmpeg |
|
| 211 |
+ .input(args.input) |
|
| 212 |
+ .audio |
|
| 213 |
+ # This works badly for some reason. |
|
| 214 |
+ .filter( |
|
| 215 |
+ 'atrim', |
|
| 216 |
+ start=f"{args.start_frame/args.input_fps}s",
|
|
| 217 |
+ end=f"{args.end_frame/args.input_fps}s",
|
|
| 218 |
+ ) |
|
| 219 |
+ ] |
|
| 220 |
+ if not args.no_audio else [] |
|
| 221 |
+ ) |
|
| 222 |
+ input_stream = ( |
|
| 223 |
+ ffmpeg |
|
| 224 |
+ .input(args.input) |
|
| 225 |
+ .output('pipe:', **pipe_args)
|
|
| 226 |
+ .global_args('-loglevel', 'error')
|
|
| 227 |
+ # .global_args('-stats')
|
|
| 228 |
+ .run_async(pipe_stdout=True) |
|
| 229 |
+ ) |
|
| 230 |
+ if not args.no_write: |
|
| 231 |
+ if args.debug: |
|
| 232 |
+ debug_stream = ( |
|
| 233 |
+ ffmpeg |
|
| 234 |
+ .input('pipe:', **pipe_args, s=debug_size, r=args.output_fps)
|
|
| 235 |
+ .output( |
|
| 236 |
+ *audio_args, args.debug, **output_args, r=args.output_fps) |
|
| 237 |
+ .global_args('-loglevel', 'error')
|
|
| 238 |
+ .run_async(pipe_stdin=True, overwrite_output=args.overwrite) |
|
| 239 |
+ ) |
|
| 240 |
+ if args.output: |
|
| 241 |
+ output_stream = ( |
|
| 242 |
+ ffmpeg |
|
| 243 |
+ .input('pipe:', **pipe_args, s=output_size, r=args.output_fps)
|
|
| 244 |
+ .output( |
|
| 245 |
+ *audio_args, args.output, **output_args, r=args.output_fps) |
|
| 246 |
+ .global_args('-loglevel', 'error')
|
|
| 247 |
+ .run_async(pipe_stdin=True, overwrite_output=args.overwrite) |
|
| 248 |
+ ) |
|
| 249 |
+ |
|
| 250 |
+ # Set up signal handler. |
|
| 251 |
+ sigint = False |
|
| 252 |
+ |
|
| 253 |
+ def sigint_handler(signum, frame): |
|
| 254 |
+ nonlocal sigint |
|
| 255 |
+ sigint = True |
|
| 256 |
+ |
|
| 257 |
+ signal.signal(signal.SIGINT, sigint_handler) |
|
| 258 |
+ |
|
| 259 |
+ # Call init. |
|
| 260 |
+ if init is not None: |
|
| 261 |
+ state = init(args) |
|
| 262 |
+ else: |
|
| 263 |
+ state = None |
|
| 264 |
+ |
|
| 265 |
+ # Process. |
|
| 266 |
+ try: |
|
| 267 |
+ for frame_num in range(args.frame_count): |
|
| 268 |
+ # Check for end frame. |
|
| 269 |
+ if frame_num >= args.end_frame: |
|
| 270 |
+ break |
|
| 271 |
+ |
|
| 272 |
+ # Gather and print info. |
|
| 273 |
+ time_now = datetime.timedelta( |
|
| 274 |
+ seconds=int(frame_num / args.input_fps)) |
|
| 275 |
+ time_duration = datetime.timedelta( |
|
| 276 |
+ seconds=int(args.duration)) |
|
| 277 |
+ elapsed_frames = f"{frame_num} / {args.frame_count-1}"
|
|
| 278 |
+ elapsed_time = f"{time_now} / {time_duration}"
|
|
| 279 |
+ argv = sys.argv |
|
| 280 |
+ if len(argv) >= 1: |
|
| 281 |
+ argv[0] = os.path.basename(argv[0]) |
|
| 282 |
+ argv = " ".join(map(shlex.quote, argv)) |
|
| 283 |
+ sys.stdout.write(f"{elapsed_frames} ({elapsed_time})\r")
|
|
| 284 |
+ |
|
| 285 |
+ # Read input. |
|
| 286 |
+ if sigint: |
|
| 287 |
+ break |
|
| 288 |
+ input_bytes = input_stream.stdout.read( |
|
| 289 |
+ args.input_width * args.input_height * 3 |
|
| 290 |
+ ) |
|
| 291 |
+ if not input_bytes: |
|
| 292 |
+ break |
|
| 293 |
+ input_frame = ( |
|
| 294 |
+ np |
|
| 295 |
+ .frombuffer(input_bytes, np.uint8) |
|
| 296 |
+ .reshape([args.input_height, args.input_width, 3]) |
|
| 297 |
+ ) |
|
| 298 |
+ |
|
| 299 |
+ # Check for start frame. |
|
| 300 |
+ if frame_num < args.start_frame: |
|
| 301 |
+ continue |
|
| 302 |
+ |
|
| 303 |
+ # Check for decimate frame. |
|
| 304 |
+ if (frame_num - args.start_frame) % args.decimate != 0: |
|
| 305 |
+ continue |
|
| 306 |
+ |
|
| 307 |
+ # Resize to working size. |
|
| 308 |
+ frame = resize( |
|
| 309 |
+ input_frame, |
|
| 310 |
+ args.working_width, |
|
| 311 |
+ args.working_height, |
|
| 312 |
+ PIL.Image.NEAREST, |
|
| 313 |
+ ) |
|
| 314 |
+ |
|
| 315 |
+ # Call process. |
|
| 316 |
+ output_frame, debug_frame = process(args, state, frame, frame_num) |
|
| 317 |
+ |
|
| 318 |
+ # Show info. |
|
| 319 |
+ if args.debug: |
|
| 320 |
+ debug_frame = text( |
|
| 321 |
+ debug_frame, |
|
| 322 |
+ elapsed_frames, |
|
| 323 |
+ 0, |
|
| 324 |
+ 0, |
|
| 325 |
+ args.thickness, |
|
| 326 |
+ ) |
|
| 327 |
+ debug_frame = text( |
|
| 328 |
+ debug_frame, |
|
| 329 |
+ elapsed_time, |
|
| 330 |
+ args.working_width, |
|
| 331 |
+ 0, |
|
| 332 |
+ args.thickness, |
|
| 333 |
+ ) |
|
| 334 |
+ debug_frame = text( |
|
| 335 |
+ debug_frame, |
|
| 336 |
+ argv, |
|
| 337 |
+ 0, |
|
| 338 |
+ args.working_height, |
|
| 339 |
+ args.thickness, |
|
| 340 |
+ ) |
|
| 341 |
+ |
|
| 342 |
+ # Show preview windows. |
|
| 343 |
+ if not args.no_preview: |
|
| 344 |
+ if args.debug: |
|
| 345 |
+ cv2.imshow( |
|
| 346 |
+ f"{args.debug}",
|
|
| 347 |
+ cv2.cvtColor(debug_frame, cv2.COLOR_RGB2BGR), |
|
| 348 |
+ ) |
|
| 349 |
+ if args.output: |
|
| 350 |
+ cv2.imshow( |
|
| 351 |
+ f"{args.output}",
|
|
| 352 |
+ cv2.cvtColor(output_frame, cv2.COLOR_RGB2BGR), |
|
| 353 |
+ ) |
|
| 354 |
+ if cv2.waitKey(1) in (ord('q'), 27):
|
|
| 355 |
+ break |
|
| 356 |
+ |
|
| 357 |
+ # Write files. |
|
| 358 |
+ if not args.no_write: |
|
| 359 |
+ if sigint: |
|
| 360 |
+ break |
|
| 361 |
+ try: |
|
| 362 |
+ if args.debug: |
|
| 363 |
+ debug_stream.stdin.write( |
|
| 364 |
+ debug_frame |
|
| 365 |
+ .astype(np.uint8) |
|
| 366 |
+ .tobytes() |
|
| 367 |
+ ) |
|
| 368 |
+ if args.output: |
|
| 369 |
+ output_stream.stdin.write( |
|
| 370 |
+ output_frame |
|
| 371 |
+ .astype(np.uint8) |
|
| 372 |
+ .tobytes() |
|
| 373 |
+ ) |
|
| 374 |
+ except BrokenPipeError: |
|
| 375 |
+ # FFmpeg has probably written some error message to stderr, |
|
| 376 |
+ # so just break. |
|
| 377 |
+ break |
|
| 378 |
+ except KeyboardInterrupt: |
|
| 379 |
+ pass |
|
| 380 |
+ finally: |
|
| 381 |
+ print("")
|
|
| 382 |
+ # Close and wait. |
|
| 383 |
+ if not args.no_write: |
|
| 384 |
+ if args.debug: |
|
| 385 |
+ debug_stream.stdin.close() |
|
| 386 |
+ debug_stream.wait() |
|
| 387 |
+ if args.output: |
|
| 388 |
+ output_stream.stdin.close() |
|
| 389 |
+ output_stream.wait() |
|
| 390 |
+ input_stream.send_signal(signal.SIGINT) |
|
| 391 |
+ input_stream.communicate() |
git.rcrnstn.net