1. audio and video fingerprinting
John Schavemaker, Werner Bailer, Peter-Jan Doets, Jaap Blom
2. techniek even in kort:
duplicaatherkenning (video fingerprinting)
• bestaat een video in onze databases?
categorisatie
• wat voor categorie video is het? Nieuws, sport, film?
object- en logoherkenning
• bestaat een object of logo (plaatje) in onze databases?
Zie ook ons online rapport over stand van de techniek:
http://research.imagesforthefuture.org/index.php/video-fingerprinting-state-of-the-art-report/
2 audio and video fingerprinting
3. duplicaatherkenning
VRAAG: bestaat een video in onze databases?
video fingerprints houden rekening
met veranderingen in:
• resolutie
• codec
• ruis
• kleur
3 audio and video fingerprinting
4. SWOT video fingerprinting
STRENGTHS WEAKNESSES
• uitontwikkelde technologie • veel concurrerende partijen, welk
• zeer goede performance op softwarepakket te kiezen?
geproduceerd materiaal • geschiktheid voor video materiaal dat
• veel commerciële pakketten niet geproduceerd is?
verkrijgbaar op de markt
OPPORTUNITIES THREATS
• grotere video databases • video fingerprints gesloten
• niet geproduceerd materiaal standaarden
• open standaard video fingerprints • versleuteling video
• combinatie met audio • slimme “gebruikers”
4 audio and video fingerprinting
5. video categorisatie
VRAAG: Wat voor categorie video is het?
Close-up gezicht, binnensport, buitensport?
images UvA
http://www.science.uva.nl/research/mediamill/
5 audio and video fingerprinting
6. SWOT video categorisatie
STRENGTHS WEAKNESSES
• veel belovende techniek • onvolwassen techniek
• generieke herkenning mogelijk • performance (sterk) afhankelijk
• aanvulling op duplicaat- en van gebruikte leervoorbeelden
objectherkenning • leren systeem voor nieuwe
• brug van de ‘semantic gap’ categorieën duurt relatief lang
OPPORTUNITIES THREATS
• combinatie van categorieën • variëteit te groot voor categorie
• sneller en beter leren • keuze van categorieën
• automatische annotatie • afhankelijk van annotatie
leervoorbeelden
6 audio and video fingerprinting
7. object- en logoherkenning
VRAAG: bestaat
een object of logo
in onze databases?
picture from http://www.omniperception.com/
7 audio and video fingerprinting
8. SWOT object- en logoherkenning
STRENGTHS WEAKNESSES
• goede, robuuste performance • alleen 2D objecten (logo’s)
• commerciële pakketten • echte duplicaatherkenning
• snel leren en herkennen • rekenintensief
• revolutie in computer vision
OPPORTUNITIES THREATS
• grotere video databases • pre-processing al het materiaal
• open standaard noodzakelijk
• 3D object herkenning • patenten
8 audio and video fingerprinting
10. Use of FP: identification
Audio/visual Fingerprint
Labeled signal extraction Fingerprints
Multimedia and
items Metadata
Metadata
Training phase
Identification phase
Unlabeled Fingerprint
Audio/visual Match Which item?
Multimedia extraction
signal Metadata
items
10 audio and video fingerprinting
11. Sound & Vision Pilot
• Observations
• Problem harder than expected
• Transformations
• Crop & scale
• Brightness/contrast
• Logos, captions
• very difficult PIP
• many matching sequences of black frames
11 audio and video fingerprinting
12. Sound & Vision Pilot – results ZiuZ
• TNO has used the ZiuZ video fingerprinting tool on the dataset
• ZiuZ video fingerprinting is optimized for child-abuse material:
• short clips
• low resolution
• low image quality
• Preliminary results on the Sound & Vision dataset show
• material is very challenging
• some but limited recall performance
• application domain differs
• queries containing multiple clips of reference material were
not enabled by this version of the tool
12 audio and video fingerprinting
13. Sound & Vision Pilot – Results JRS
• Recall: 36% (min: 16%, max. 55%)
• Precision: difficult to determine, many black
sequences matching, needs manual checking
13 audio and video fingerprinting
14. Sound & Vision Pilot - Results
• Transformations our system handles
14 audio and video fingerprinting
15. Sound & Vision Pilot - Results
• False positives
15 audio and video fingerprinting
16. Experiments with SIFT (1)
• we do not have a SIFT based fingerprinting
solution in the consortium
• JRS has SIFT-based interactive tool to locate
recurring objects in video
• created video from episode + source clips and
performed analysis and search
16 audio and video fingerprinting
19. Experiments with SIFT (4)
• Conclusion
• SIFT can handle cases of scaling and cropping
reliably
• even PIP with distortions
• Scalability issues
• time for extraction and esp. matching
• not sure if ranking of matches is still reliable on
huge datasets
19 audio and video fingerprinting
20. Characteristics of the data set - audio
• Not all archive fragments contain audio
• Often the original audio is used – just cut-and-paste, no serious
distortions
• Sometimes the audio is replaced or combined with a voice over
• Time segmentation of the audio in the episode is different from
the video used. The audio is not always used with the
corresponding video fragments. Example on next slide illustrates
this. The other ways around, and other variations also occur.
20 audio and video fingerprinting
21. Characteristics of the data set – audio example
Time line of one
archive video
video
audio
Time line of one
Andere Tijden episode
video
audio
Continuous audio fragment, with several shorter video fragments
21 audio and video fingerprinting
22. Characteristics of the data set - audio
• Limitations of the use of audio
• the reference material must contain audio
• the audio track might not originate from the same material as
the video track; this is dependent on the video material used.
• the playout speed must not be changed too much (less than
+/- 2%)
• Advantages of the use of audio
• Highly robust algorithms
• Usually audio is undistorted; video is cropped, scaled, etc.
• Audio usually is used continuously, while video fragments are
cut-and-paste from different sections of the reference video,
and ‘glued together’.
22 audio and video fingerprinting
23. Identification results - audio
• Only checked if the correct archive file name is returned
False
Episode Correct Missed Positive
Liggadjati 8 3 0
Veertig jaar STER-reclame 10 4 1
75 jaar afsluitdijk 0 5 2
Strijd tegen de file 9 1 6
Kronkels van de Maas 1 9 1
Op zoek naar Nederland 2 6 1
Modderen in de polder: Lelystad 3 1 2
Burgemeesters in oorlogstijd 6 10 0
De wording van Paars 8 1 0
Pim en zijn volk 7 3 0
23 audio and video fingerprinting
silent parts in the video
24. Fingerprinting – audio algorithm
• Algorithm well-known from literature:
• Haitsma, Kalker, “A Highly Robust Audio Fingerprinting
System”, In Proceedings of 3rd International Conference
onMusic Information Retrieval (ISMIR), October 2002.
• Features: energy in 33 audio frequency bands
• Every 11.6 ms a 32-bit sub-fingerprint is computed, consisting of
coarsely quantized differences between these energy samples
• Fingerprint consists of a time series of sub-fingerprints
• The implementation returns the best matching fragments only
(settings to return no false positives)
• Algorithm is highly robust, and highly discriminative
24 audio and video fingerprinting
25. Future improvements on current results
• Trailing parts contain silence and black frames (no content). The
silences give rise to false positives and irrelevant detections. A
silence/activity detector is needed to exclude these parts.
• Our current implementation from literature allows for only one
fragment per reference file to be returned.
• Our current implementation has only coarse time localization.
• Combination of audio and video fingerprinting
25 audio and video fingerprinting
26. Consortium
http://instituut.beeldengeluid.nl/
http://www.joanneum.at/en/digital.html
http://www.ziuz.com
http://hs-art.com/
http://www.tno.nl
26 audio and video fingerprinting