The 7 Things I Know About Cyber Security After 25 Years | April 2024
Pycvf
1. PyCVF
Overview of
the Python Computer Vision Framework
(preliminary version)
Bertrand NOUVEL
Japanese French Laboratory for Informatics
bertrand.nouvel@gmail.com
May 2010
2. Traditional Framework
Machine
Create Compute Extract Compute Fusion
Learning
Video DB Shots Keyframes Features Rendering
Indexing
Program 1 Program 2 Program 3 Program 4 Program 5 Program 6
Program 7
● Too many steps
● Many access to files : low performances
● Difficult to check and to run
● Not easy to perform testings
3. Models We Need to
Represent
● Learning all along the process;
● Many classspecific learners;
● Link with highlevel ontologies
● Ability to provide topdown and bottomup
communications
Too complex fort traditional frameworks !
→ Need for improvement
4. Can We Do Better ?
● Avoid usage of archaic file formats
● Use better software design
● Inspire from what is meant to be ”easy”
Blender OpenCV
PureData
5. PyCVF is a Framework
OPENCV WEKA
Dependencies
Libraries MPT Orange
Toolkits
(features) Marsyas R
Framework
PyCVF
(uniformize concepts)
Applications
Analyze Mosaicing
Synthetize Indexer Recognizer
Textures Loop generator
(use and
extend the framework)
7. PyCVF Datatypes
● Images → 3d Numpy Array
● Video → Video Player Object (play/pause/seek)
● Audio → Audio Player Object (play/pause/seek)
● Vectors → 1d Numpy Array
● Vectorset → 2d – Numpy Array
Datatypes can solve the following questions ?
● How to store the data ?
● How to display the data ?
8. PyCVF Databases
Supported operations
● Get Datatype
● List Items
DB
● Query
● List Keys
● List Labels
● Query Label
10. Example Meta databases
● Limit : Limit the number of elements
● Transformed : Apply a transformation
● Exploded : Obtain subelements
● TrainDB : Create a training associated with a seed
● TestDB : Create a test set associated with a seed
● Randomized : Shuffle the elements of a database
● Agglomerated : Merge databases together
● As_once : Return databases as elements
21. Making Experiments
● Easy sampling of
parameter space
● Automatic
parallelization of
runs
● Specific graphics
● Specific feature
● Minimization
Performances of PyCVF of index queries
23. Browsing Databases
● Onthefly
transformations
● Easy queries to already
computed index
24. Viewing Element 1 by 1
● Onthefly
transformations
● Easy queries to already
computed index
25. Revision Plan
0.2 0.3
● Do call for volunteers ● Add Features
● Ensure global design ● Improve feature
● Easier install stability
● Increase corerobustness
● Provide exhaustive set
of basic features
● Regression tests
● Start add new features
27. Included Wrappers
● PyFFMPEG is able to
PyFFMPEG play in realtime 121
videos at the same
time.
28. Included Wrappers
Implements
●
OpenCV-Cython exhaustively C API
● Numpy compatibility
● Accessible pointers
● May be deprecated due
to new C++ wrappers
29. Included Wrappers
● Fast nearest neighbor
PySASH search
● Patented + NOT GPL
30. Included Wrappers
● Fast nearest neighbor
PyLSH search with LSH
● Different
implementations
● Still Buggy
● Contributions are
welcome