video_source
VideoConsumer
¶
This class should be consumed as part of internal implementation. It provides abstraction around stream consumption strategies.
Source code in inference/core/interfaces/camera/video_source.py
723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 |
|
VideoSource
¶
Source code in inference/core/interfaces/camera/video_source.py
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 |
|
__next__()
¶
Method allowing to use VideoSource
convenient to read frames
Returns: VideoFrame
Example
source = VideoSource.init(video_reference="./some.mp4")
source.start()
for frame in source:
pass
Source code in inference/core/interfaces/camera/video_source.py
702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 |
|
frame_ready()
¶
Method to check if decoded frame is ready for consumer
Returns: boolean flag indicating frame readiness
Source code in inference/core/interfaces/camera/video_source.py
503 504 505 506 507 508 509 |
|
get_state()
¶
Method to get current state of the VideoSource
Returns: StreamState
Source code in inference/core/interfaces/camera/video_source.py
495 496 497 498 499 500 501 |
|
init(video_reference, buffer_size=DEFAULT_BUFFER_SIZE, status_update_handlers=None, buffer_filling_strategy=None, buffer_consumption_strategy=None, adaptive_mode_stream_pace_tolerance=DEFAULT_ADAPTIVE_MODE_STREAM_PACE_TOLERANCE, adaptive_mode_reader_pace_tolerance=DEFAULT_ADAPTIVE_MODE_READER_PACE_TOLERANCE, minimum_adaptive_mode_samples=DEFAULT_MINIMUM_ADAPTIVE_MODE_SAMPLES, maximum_adaptive_frames_dropped_in_row=DEFAULT_MAXIMUM_ADAPTIVE_FRAMES_DROPPED_IN_ROW, video_source_properties=None, source_id=None)
classmethod
¶
This class is meant to represent abstraction over video sources - both video files and
on-line streams that are possible to be consumed and used by other components of inference
library.
Before digging into details of the class behaviour, it is advised to familiarise with the following concepts and implementation assumptions:
- Video file can be accessed from local (or remote) storage by the consumer in a pace dictated by its processing capabilities. If processing is faster than the frame rate of video, operations may be executed in a time shorter than the time of video playback. In the opposite case - consumer may freely decode and process frames in its own pace, without risk for failures due to temporal dependencies of processing - this is classical offline processing example.
- Video streams, on the other hand, usually need to be consumed in a pace near to their frame-rate - in other words - this is on-line processing example. Consumer being faster than incoming stream frames cannot utilise its resources to the full extent as not-yet-delivered data would be needed. Slow consumer, however, may not be able to process everything on time and to keep up with the pace of stream - some frames would need to be dropped. Otherwise - over time, consumer could go out of sync with the stream causing decoding failures or unpredictable behavior.
To fit those two types of video sources, VideoSource
introduces the concept of buffered decoding of
video stream (like at the YouTube - player buffers some frames that are soon to be displayed).
The way on how buffer is filled and consumed dictates the behavior of VideoSource
.
Starting from BufferFillingStrategy
- we have 3 basic options:
* WAIT: in case of slow video consumption, when buffer is full - VideoSource
will wait for
the empty spot in buffer before next frame will be processed - this is suitable in cases when
we want to ensure EACH FRAME of the video to be processed
* DROP_OLDEST: when buffer is full, the frame that sits there for the longest time will be dropped -
this is suitable for cases when we want to process the most recent frames possible
* DROP_LATEST: when buffer is full, the newly decoded frame is dropped - useful in cases when
it is expected to have processing performance drops, but we would like to consume portions of
video that are locally smooth - but this is probably the least common use-case.
On top of that - there are two ADAPTIVE strategies: ADAPTIVE_DROP_OLDEST and ADAPTIVE_DROP_LATEST, which are equivalent to DROP_OLDEST and DROP_LATEST with adaptive decoding feature enabled. The notion of that mode will be described later.
Naturally, decoded frames must also be consumed. VideoSource
provides a handy interface for reading
a video source frames by a SINGLE consumer. Consumption strategy can also be dictated via
BufferConsumptionStrategy
:
* LAZY - consume all the frames from decoding buffer one-by-one
* EAGER - at each readout - take all frames already buffered, drop all of them apart from the most recent
In consequence - there are various combinations of BufferFillingStrategy
and BufferConsumptionStrategy
.
The most popular would be:
* BufferFillingStrategy.WAIT
and BufferConsumptionStrategy.LAZY
- to always decode and process each and
every frame of the source (useful while processing video files - and default behaviour enforced by
inference
if there is no explicit configuration)
* BufferFillingStrategy.DROP_OLDEST
and BufferConsumptionStrategy.EAGER
- to always process the most
recent frames of source (useful while processing video streams when low latency [real-time experience]
is required - ADAPTIVE version of this is default for streams)
ADAPTIVE strategies were introduced to handle corner-cases, when consumer hardware is not capable to consume
video stream and process frames at the same time (for instance - Nvidia Jetson devices running processing
against hi-res streams with high FPS ratio). It acts with buffer in nearly the same way as DROP_OLDEST
and DROP_LATEST
strategies, but there are two more conditions that may influence frame drop:
* announced rate of source - which in fact dictate the pace of frames grabbing from incoming stream that
MUST be met by consumer to avoid strange decoding issues causing decoder to fail - if the pace of frame grabbing
deviates too much - decoding will be postponed, and frames dropped to grab next ones sooner
* consumption rate - in resource constraints environment, not only decoding is problematic from the performance
perspective - but also heavy processing. If consumer is not quick enough - allocating more useful resources
for decoding frames that may never be processed is a waste. That's why - if decoding happens more frequently
than consumption of frame - ADAPTIVE mode causes decoding to be done in a slower pace and more frames are just
grabbed and dropped on the floor.
ADAPTIVE mode increases latency slightly, but may be the only way to operate in some cases.
Behaviour of adaptive mode, including the maximum acceptable deviations of frames grabbing pace from source,
reader pace and maximum number of consecutive frames dropped in ADAPTIVE mode are configurable by clients,
with reasonable defaults being set.
VideoSource
emits events regarding its activity - which can be intercepted by custom handlers. Take
into account that they are always executed in context of thread invoking them (and should be fast to complete,
otherwise may block the flow of stream consumption). All errors raised will be emitted as logger warnings only.
VideoSource
implementation is naturally multithreading, with different thread decoding video and different
one consuming it and manipulating source state. Implementation of user interface is thread-safe, although
stream it is meant to be consumed by a single thread only.
ENV variables involved: * VIDEO_SOURCE_BUFFER_SIZE - default: 64 * VIDEO_SOURCE_ADAPTIVE_MODE_STREAM_PACE_TOLERANCE - default: 0.1 * VIDEO_SOURCE_ADAPTIVE_MODE_READER_PACE_TOLERANCE - default: 5.0 * VIDEO_SOURCE_MINIMUM_ADAPTIVE_MODE_SAMPLES - default: 10 * VIDEO_SOURCE_MAXIMUM_ADAPTIVE_FRAMES_DROPPED_IN_ROW - default: 16
As an inference
user, please use .init() method instead of constructor to instantiate objects.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
video_reference |
Union[str, int]
|
Either str with file or stream reference, or int representing device ID |
required |
buffer_size |
int
|
size of decoding buffer |
DEFAULT_BUFFER_SIZE
|
status_update_handlers |
Optional[List[Callable[[StatusUpdate], None]]]
|
List of handlers for status updates |
None
|
buffer_filling_strategy |
Optional[BufferFillingStrategy]
|
Settings for buffer filling strategy - if not given - automatic choice regarding source type will be applied |
None
|
buffer_consumption_strategy |
Optional[BufferConsumptionStrategy]
|
Settings for buffer consumption strategy, if not given - automatic choice regarding source type will be applied |
None
|
adaptive_mode_stream_pace_tolerance |
float
|
Maximum deviation between frames grabbing pace and stream pace that will not trigger adaptive mode frame drop |
DEFAULT_ADAPTIVE_MODE_STREAM_PACE_TOLERANCE
|
adaptive_mode_reader_pace_tolerance |
float
|
Maximum deviation between decoding pace and stream consumption pace that will not trigger adaptive mode frame drop |
DEFAULT_ADAPTIVE_MODE_READER_PACE_TOLERANCE
|
minimum_adaptive_mode_samples |
int
|
Minimal number of frames to be used to establish actual pace of processing, before adaptive mode can drop any frame |
DEFAULT_MINIMUM_ADAPTIVE_MODE_SAMPLES
|
maximum_adaptive_frames_dropped_in_row |
int
|
Maximum number of frames dropped in row due to application of adaptive strategy |
DEFAULT_MAXIMUM_ADAPTIVE_FRAMES_DROPPED_IN_ROW
|
video_source_properties |
Optional[dict[str, float]]
|
Optional dictionary with video source properties corresponding to OpenCV VideoCapture properties cv2.CAP_PROP_* to set values for the video source. |
None
|
source_id |
Optional[int]
|
Optional identifier of video source - mainly useful to recognise specific source when multiple ones are in use. Identifier will be added to emitted frames and updates. It is advised to keep it unique within all sources in use. |
None
|
Returns: Instance of VideoSource
class
Source code in inference/core/interfaces/camera/video_source.py
169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 |
|
mute()
¶
Method to be used to mute source consumption. Muting is an equivalent of pause for stream - where frames grabbing is not put on hold, just new frames decoding and buffering is not allowed - causing intermediate frames to be dropped. May be also used against files, although arguably less useful. Eligible to be used in states: [RUNNING] End state: * MUTED
Thread safe - only one transition of states possible at the time.
Throws: * StreamOperationNotAllowedError: if executed in context of incorrect state of the source
Source code in inference/core/interfaces/camera/video_source.py
452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 |
|
pause()
¶
Method to be used to pause source consumption. During pause - no new frames are consumed. Used on on-line streams for too long may cause stream disconnection. Eligible to be used in states: [RUNNING] End state: * PAUSED
Thread safe - only one transition of states possible at the time.
Throws: * StreamOperationNotAllowedError: if executed in context of incorrect state of the source
Source code in inference/core/interfaces/camera/video_source.py
430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 |
|
read_frame(timeout=None)
¶
Method to be used by the consumer to get decoded source frame.
Throws: * EndOfStreamError: when trying to get the frame from closed source.
Source code in inference/core/interfaces/camera/video_source.py
511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 |
|
restart(wait_on_frames_consumption=True, purge_frames_buffer=False)
¶
Method to restart source consumption. Eligible to be used in states: [MUTED, RUNNING, PAUSED, ENDED, ERROR]. End state: * INITIALISING - that should change into RUNNING once first frame is ready to be grabbed * ERROR - if it was not possible to connect with source
Thread safe - only one transition of states possible at the time.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
wait_on_frames_consumption |
bool
|
Flag telling if all frames from buffer must be consumed before completion of this operation. |
True
|
Throws: * StreamOperationNotAllowedError: if executed in context of incorrect state of the source * SourceConnectionError: if source cannot be connected
Source code in inference/core/interfaces/camera/video_source.py
345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 |
|
resume()
¶
Method to recover from pause or mute into running state. [PAUSED, MUTED] End state: * RUNNING
Thread safe - only one transition of states possible at the time.
Throws: * StreamOperationNotAllowedError: if executed in context of incorrect state of the source
Source code in inference/core/interfaces/camera/video_source.py
475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 |
|
start()
¶
Method to be used to start source consumption. Eligible to be used in states: [NOT_STARTED, ENDED, (RESTARTING - which is internal state only)] End state: * INITIALISING - that should change into RUNNING once first frame is ready to be grabbed * ERROR - if it was not possible to connect with source
Thread safe - only one transition of states possible at the time.
Throws: * StreamOperationNotAllowedError: if executed in context of incorrect state of the source * SourceConnectionError: if source cannot be connected
Source code in inference/core/interfaces/camera/video_source.py
376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 |
|
terminate(wait_on_frames_consumption=True, purge_frames_buffer=False)
¶
Method to be used to terminate source consumption. Eligible to be used in states: [MUTED, RUNNING, PAUSED, ENDED, ERROR, (RESTARTING - which is internal state only)] End state: * ENDED - indicating success of the process * ERROR - if error with processing occurred
Must be used to properly dispose resources at the end.
Thread safe - only one transition of states possible at the time.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
wait_on_frames_consumption |
bool
|
Flag telling if all frames from buffer must be consumed before completion of this operation. |
True
|
Throws: * StreamOperationNotAllowedError: if executed in context of incorrect state of the source
Source code in inference/core/interfaces/camera/video_source.py
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 |
|
get_from_queue(queue, timeout=None, on_successful_read=lambda: None, purge=False)
¶
Function is supposed to take element from the queue waiting on the first element to appear using timeout
parameter. One may ask to go to the very last element of the queue and return it - then purge
should be set
to True. No additional wait on new elements to appear happen and the purge stops once queue is free returning last
element consumed.
queue.task_done() and on_successful_read(...) will be called on each received element.
Source code in inference/core/interfaces/camera/video_source.py
993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 |
|