PySceneDetect Releases
PySceneDetect 0.5
0.5.6.1 (October 11, 2021)
- Fix crash when using
detect-content
ordetect-adaptive
with latest version of OpenCV (thanks @bilde2910)
0.5.6 (August 15, 2021)
Release Notes
- New detection algorithm:
detect-adaptive
which works similar todetect-content
, but with reduced false negatives during fast camera movement (thanks @scarwire and @wjs018) - Images generated by
save-images
can now be resized via the command line - Statsfiles now work properly with
detect-threshold
- Removed the
-p
/--min-percent
option fromdetect-threshold
- Add new option
-l
/--luma-only
todetect-content
/detect-adaptive
to only consider brightness channel (useful for greyscale videos)
Changelog
- [feature] New adaptive content detector algorithm
detect-adaptive
(#153, thanks @scarwire and @wjs018) - [feature] Images generated with the
save-images
command (scene_manager.save_images()
function in the Python API) can now be scaled or resized (#160 and PR #203, thanks @wjs018)- Images can be resized by a constant scaling factory using
-s
/--scale
(e.g.--scale 0.5
shrinks the height/width by half) - Images can be resized to a specified height (
-h
/--height
) and/or width (-w
/--width
), in pixels; if only one is specified, the aspect ratio of the original video is kept
- Images can be resized by a constant scaling factory using
- [api] Calling
seek()
on aVideoManager
will now respect the end time if set - [api] The
split_video_
functions now return the exit code of invokingffmpeg
ormkvmerge
(#209, thanks @AdrienLF) - [api] Removed the
min_percent
argument fromThresholdDetector
as was not providing any performance benefit for the majority of use cases (#178) - [bugfix] The
detect-threshold
command now works properly with a statsfile (#211, thanks @jeremymeyers) - [bugfix] Fixed crash due to unhandled
TypeError
exception when using non-PyPI OpenCV packages from certain Linux distributions (#220) - [bugfix] A warning is now displayed for videos which may not be decoded correctly, esp. VP9 (#86, thanks @wjs018)
- [api] A named logger is now used for both API and CLI logging instead of the root logger (#205)
Known Issues
- Image sequences or URL inputs are not supported by the
save-images
orsplit-video
commands - Variable framerate videos (VFR) are not fully supported, and will yield incorrect timestamps (#168)
0.5.5 (January 17, 2021)
Release Notes
- One of the last major updates before transitioning to the new v0.6.x API
- The
--min-scene-len
/-m
option is now global rather than per-detector - There is a new global option
--drop-short-scenes
to go along with-m
- Removed first row from statsfiles so it is a valid CSV file
- The progress bar now correctly resizes when the terminal is resized
- Image sequences and URLs are now supported for input via the CLI/API
- Images exported using the
save-images
command are now resized to match the display aspect ratio - A new flag
-s
/--skip-cuts
has been added to thelist-scenes
command to allow standardized processing - The functionality of
save-images
is now accessible via the Python API through thesave_images()
function inscenedetect.scene_manager
- Under the
save-images
command, renamed--image-frame-margin
to--frame-margin
, added short option-m
, and increased the default value from 0 to 1 due to instances of the last frame of a video being occasionally missed (set-m 0
to restore original behaviour)
Changelog
- [bugfix] Allow image sequences and URLs to be used as inputs (#152 and #188)
- [bugfix] Pixel aspect ratio is now applied when using
save-images
(#195) - [cli] Renamed
--image-frame-margin
to--frame-margin
insave-images
command, added short option-m
as alias - [bugfix] Fix
save-images
command not saving the last frame by modifying seeking, as well as increasing default of--frame-margin
from 0 to 1 - [cli] Make
--min-scene-len
a global option rather than per-detector (#131, thanks @tonycpsu) - [feature] Added
--drop-short-scenes
option to remove all scenes smaller than--min-scene-len
, instead of merging them - [cli] Add
-s
/--skip-cuts
option tolist-scenes
command to allow outputting a scene list CSV file as compliant with RFC 4180 (#136) - [enhancement] Removed first row from statsfile to comply with RFC 4180, includes backwards compatibility so existing statsfiles can still be loaded (#136)
- [api] Add argument
include_cut_list
towrite_scene_list
method inSceneManager
to support #136 - [api] Removed unused argument base_timecode from
StatsManager.load_from_csv()
method - [api] Make the
base_timecode
argument optional on theSceneManager
methodsget_scene_list()
,get_cut_list()
, andget_event_list()
(#173) - [api] Support for live video stream callbacks by adding new
callback
argument to thedetect_scenes()
method ofSceneManager
(#5, thanks @mhashim6) - [bugfix] Fix unhandled exception causing improper error message when a video fails to load on non-Windows platforms (#192)
- [enhancement] Enabled dynamic resizing for progress bar (#193)
- [enhancement] Always ouptut version number via logger to assist with debugging (#171)
- [bugfix] Resolve RuntimeWarning when running as module (#181)
- [api] Add
save_images()
function toscenedetect.scene_manager
module which exposes the same functionality as the CLIsave-images
command (#88) - [api] Removed
close_captures()
andrelease_captures()
functions fromscenedetect.video_manager
module
Known Issues
- Certain non-PyPI OpenCV packages may cause a crash with the message
TypeError: isinstance() arg 2 must be a type or tuple of types
- as a workaround, install the Python OpenCV package by runningpip install scenedetect[opencv]
(#220) - Image sequences or URL inputs are not supported by the
save-images
orsplit-video
commands - Variable framerate videos (VFR) are not fully supported, and will yield incorrect timestamps (#168)
0.5.4 (September 14, 2020)
Release Notes
- Improved performance when using
time
andsave-images
commands - Improved performance of
detect-threshold
when using a small minimum percent - Fix crash when using
detect-threshold
with a statsfile - Fix crash when using
save-images
command under Python 2.7 - Support for Python 3.3 and 3.4 has been deprecated (see below)
Changelog
- [bugfix] fix
detect-threshold
crash when using statsfile (#122) - [bugfix] fix
save-images
command under Python 2.7 (#174, thanks @santiagodemierre) - [bugfix] gracefully exit and show link to FAQ when number of scenes is too large to split with mkvmerge on Windows (see #164
- [enhancement] Improved seeking performance, greatly improves performance of the
time
andsave-images
commands (#98 and PR #163 - thanks @obroomhall) - [enhancement] improve
detect-threshold
performance when min-percent is less than 50% - [bugfix] Fixed issue where video loading would fail silently due to multiple audio tracks (#179)
- [general] Made
tqdm
a regular requirement and not an extra (#180) - [general] Support for Python 3.3 and 3.4 has been deprecated. Newer builds may still work on these Python versions, but future releases are not tested against these versions. This decision was made as part of #180
Known Issues
- Variable framerate videos are not supported properly currently (#168), a warning may be added in the next release to indicate when a VFR video is detected, until this can be properly resolved (#168)
0.5.3 (July 12, 2020)
Release Notes
- Resolved long-standing bug where
split-video
command would duplicate certain frames at the beginning/end of the output (#93) - This was determined to be caused by copying (instead of re-encoding) the audio track, causing extra frames to be brought in when the audio samples did not line up on a frame boundary (thank you @joshcoales for your assistance)
- Default behavior is to now re-encode audio tracks using the
aac
codec when usingsplit-video
(it can be overriden in both the command line and Python interface) - Improved timestamp accuracy when using
split-video
command to further reduce instances of duplicated or off-by-one frame issues - Fixed application crash when using the
-l
/--logfile
argument
Changelog
- [bugfix] Changed default audio codec from 'copy' to 'aac' when splitting scenes with
ffmpeg
to reduce frequency of frames from next scene showing up at the end of the current one when split usingffmpeg
(see #93, #159, and PR #166 - thank you everyone for your assistance, especially joshcoales, amvscenes, jelias, and typoman). If this still occurs, please provide any information you can by filing a new issue on Github. - [enhancement]
video_splitter
module now has completed documentation - [bugfix] improve timestamp accuracy using the
split-video
command due to timecode formatting - [bugfix] fix crash when supplying
-l
/--logfile
argument (see #169, thanks @typoman)
Known Issues
- Seeking through long videos is inefficient, causing the
time
andsave-images
command to take a long time to run. This will be resolved in the next release (see #98) - The
save-images
command causes PySceneDetect to crash under Python 2.7 (see #174) - Using
detect-threshold
with a statsfile causes PySceneDetect to crash (see #122) - Variable framerate videos are not supported properly currently (#168), a warning may be added in the next release to indicate when a VFR video is detected, until this can be properly resolved (#168)
- Videos with multiple audio tracks may not work correctly, see this comment on #179 for a workaround using
ffmpeg
ormkvmerge
0.5.2 (March 29, 2020)
- [enhancement]
--min-duration
now accepts a timecode in addition to frame number (#128, thanks @tonycpsu) - [feature] Add
--image-frame-margin
option tosave-images
command to ignore a number of frames at the start/end of a scene (#129, thanks @tonycpsu) - [bugfix]
--min-scene-len
option was not respected by first scene (#105, thanks @charlesvestal) - [bugfix] Splitting videos with an analyzed duration only splits within analyzed area (#106, thanks @charlesvestal)
- [bugfix] Improper start timecode applied to the
split-video
command when usingffmpeg
(#93, thanks @typoman) - [bugfix] Added links and filename sanitation to html output (#139 and #140, thanks @wjs018)
- [bugfix] UnboundLocalError in
detect_scenes
whenframe_skip
is larger than 0 (#126, thanks @twostarxx)
0.5.1.1 (August 3, 2019)
- minor re-release of v0.5.1 which updates the setup.py file to return OpenCV as an optional dependency
- to install from pip now with all dependencies:
pip install scenedetect[opencv,progress_bar]
- to install only PySceneDetect:
pip install scenedetect
(separate OpenCV installation required) - the release notes of v0.5.1 have been modified to include the prior command
- no change to PySceneDetect program version
- [feature] add
get_duration
method to VideoManager (#109, thanks @arianaa30)
0.5.1 (July 20, 2019)
- [feature] Add new
export-html
command to the CLI (thanks @wjs018) - [bugfix] VideoManager read function failed on multiple videos (thanks @ivan23kor)
- [bugfix] Fix crash when no scenes are detected (#79, thanks @raj6996)
- [bugfix] Fixed OpenCV not getting installed due to missing dependency (#73)
- [enhance] When no scenes are detected, the whole video is now returned instead of nothing (thanks @piercus)
- Removed Windows installer due to binary packages now being available, and to streamline the release process (see #102 for more information). When you type
pip install scenedetect[opencv,progress_bar]
, all dependencies will be installed.
0.5 (August 31, 2018)
- major release, includes stable Python API with examples and updated documentation
- numerous changes to command-line interface with addition of sub-commands (see the new manual for updated usage information)
- [feature] videos are now split using
ffmpeg
by default, resulting in frame-perfect cuts (can still usemkvmerge
by specifying the-c
/--copy
argument to thesplit-video
command) - [enhance] image filename numbers are now consistent with those of split video scenes (PR #39, thanks @e271828-)
- [enhance] 5-10% improvement in processing performance due to reduced memory copy operations (PR #40, thanks @elcombato)
- [enhance] updated exception handling to raise proper standard exceptions (PR #37, thanks @talkain)
- several fixes to the documentation, including improper dates and outdated CLI arguments (PR #26 and #, thanks @elcombato, and @colelawrence)
- numerous other PRs and issues/bug reports that have been fixed - there are too many to list individually here, so I want to extend a big thank you to everyone who contributed to making this release better
- [enhance] add Sphinx-generated API documentation (available at: http://manual.scenedetect.com)
- [project] move from BSD 2-clause to 3-clause license
PySceneDetect 0.4
0.4 (January 14, 2017)
- major release, includes integrated scene splitting via mkvmerge, changes meaning of
-o
/--output
option - [feature] specifying
-o OUTPUT_FILE.mkv
will now automatically split the input video, generating a new video clip for each detected scene in sequence, starting withOUTPUT_FILE-001.mkv
- [enhance] CSV file output is now specified with the
-co
/--csv-output
option (note, used to be-o
in versions of PySceneDetect < 0.4)
PySceneDetect 0.3-beta
0.3.6 (January 12, 2017)
- [enhance] performance improvement when using
--frameskip
option (thanks @marcelluzs) - [internal] moved application state and shared objects to a consistent interface (the
SceneManager
object) to greatly reduce the number of required arguments for certain API functions - [enhance] added installer for Windows builds (64-bit only currently)
0.3.5 (August 2, 2016)
- [enhance] initial release of portable build for Windows (64-bit only), including all dependencies
- [bugfix] fix unrelated exception thrown when video could not be loaded (thanks @marcelluzs)
- [internal] fix variable name typo in API documentation
0.3.4 (February 8, 2016)
- [enhance] add scene length, in seconds, to output file (
-o
) for easier integration withffmpeg
/libav
- [enhance] improved performance of content detection mode by caching intermediate HSV frames in memory (approx. 2x faster)
- [enhance] show timecode values in terminal when using extended output (
-l
) - [feature] add fade bias option (
-fb
/--fade-bias
) to command line (threshold mode only)
0.3.3 (January 27, 2016)
- [bugfix] output scenes are now correctly written to specified output file when using -o flag (fixes #11)
- [bugfix] fix indexing exception when using multiple scene detectors and outputting statistics
- [internal] distribute package on PyPI, version move from beta to stable
- [internal] add function to convert frame number to formatted timecode
- [internal] move file and statistic output to Python
csv
module
0.3.2-beta (January 26, 2016)
- [feature] added
-si
/--save-images
flag to enable saving the first and last frames of each detected scene as an image, saved in the current working directory with the original video filename as the output prefix - [feature] added command line options for setting start and end times for processing (
-st
and-et
) - [feature] added command line option to specify maximum duration to process (
-dt
, overrides-et
)
0.3.1-beta (January 23, 2016)
- [feature] added downscaling/subsampling option (
-df
/--downscale-factor
) to improve performance on higher resolution videos - [feature] added frameskip option (
-fs
/--frame-skip
) to improve performance on high framerate videos, at expense of frame accuracy and possible inaccurate scene cut prediction - [enhance] added setup.py to allow for one-line installation (just run
python setup.py install
after downloading and extracting PySceneDetect) - [internal] additional API functions to remove requirement on passing OpenCV video objects, and allow just a file path instead
0.3-beta (January 8, 2016)
- major release, includes improved detection algorithms and complete internal code refactor
- [feature] content-aware scene detection using HSV-colourspace based algorithm (use
-d content
) - [enhance] added CLI flags to allow user changes to more algorithm properties
- [internal] re-implemented threshold-based scene detection algorithm under new interface
- [internal] major code refactor including standard detection algorithm interface and API
- [internal] remove statistics mode until update to new detection mode interface
PySceneDetect 0.2-alpha
0.2.4-alpha (December 22, 2015)
- [bugfix] updated OpenCV compatibility with self-reported version on some Linux distributions
0.2.3-alpha (August 7, 2015)
- [bugfix] updated PySceneDetect to work with latest OpenCV module (ver > 3.0)
- [bugfix] added compatibility/legacy code for older versions of OpenCV
- [feature] statsfile generation includes expanded frame metrics
0.2.2-alpha (November 25, 2014)
- [feature] added statistics mode for generating frame-by-frame analysis (-s / --statsfile flag)
- [bugfix] fixed improper timecode conversion
0.2.1-alpha (November 16, 2014)
- [enhance] proper timecode format (HH:MM:SS.nnnnn)
- [enhance] one-line of CSV timecodes added for easy splitting with external tool
0.2-alpha (June 9, 2014)
- [enhance] now provides discrete scene list (in addition to fades)
- [feature] ability to output to file (-o / --output flag)
PySceneDetect 0.1-alpha
0.1-alpha (June 8, 2014)
- first public release
- [feature] threshold-based fade in/out detection