Blender



Proposal for Blender Conference 2008

Category : papers

Author :

Alain Lioret

8 bis, rue de la Marne

95220 Herblay

France

mail : lioretalain@

Title of paper :

an Artificial Intelligence Module inside Blender for expert modeling, texturing, animating and rendering.

Summary :

The future for 3D creation softwares should include some artificial intelligence functions. None of the great 3D commercial softwares (except Endorphin from Natural Motion) proposes some tools as these described in this paper actually. Blender would be the first 3D creation suite (modeling, texturing, animation, rendering) including some modules based on optimisation techniques, machine learning and solution research, using Genetic Algorithms, Genetic Programming, Neural Networks, Self Organizing Map, Bayesian Networks, Fuzzy Logic and Principal Component Analysis.

These tools are not just for the Blender Game Engine, but also for many complex tasks as :

• Searching new textures

• Searching new 3D models

• Searching new styles for animated figures

• Machine Learning and complex set of parameters research

• Generative Animation

• Searching for new shaders

• Searching for complex parameters for rendering

• Generative Art

• etc.

Some examples of these tools are shown in my GaBuZoR system, as described in my book (in French) : « Emergence de Nouvelles Esthétiques du Mouvement » (Emergenge of new Aesthetic for Animation). (See Bibliography below).

We would make some new advanced functions in Blender (as some generics nodes) with the name « BlenderAI », and this should be a very innovative and powering framework.

The BlenderAI module should be developped for Blender 2.50 : as in , and with some new Nodes in the Node Editor.

The paper propose many examples, showing the big interest to have these kinds of tools in a 3D creation software (see also the links at the end of this document).

See these snapshots for example (from the complete paper) :

BlenderAI as a new option in the « Add Node » menu :

An example of Genetic Algorithm applied to an image search :

An example of Neural Network used to make an armature to walk :

Equipment for the presentation :

video projector

Biography :

Alain Lioret is a generative artist and educator exploring the applications of the artificial life and the artificial intelligence (cellular machines, neural networks, evolutionism). Lioret has exhibited and lectured in France (Laval Virtual 2005,2006, Ludovia 2005, Paris-Cité 1990,1991, Soreze 2004, Ecole d’Art de Rueil 2004, Espace EDF-Electra 2002, Centre International du Vitrail à Chartres 2003), Great-Britain (Digital City 2005,2006, AV Festival 2006), USA (Siggraph 2005), Italy (Generative Art 2004,2005), Latvia (PCNM 2005), Peru (ArtWare 4), Spain (EvoMusart 2007), etc.

He received his PHD in “Esthétique, Sciences et Technologies des Arts” from Paris 8 University, department “Arts et Technologies de l’Image” He is now an assistant professor in this university. He has wrote the book “Emergence de Nouvelles Esthétiques du Mouvement”, published in 2004 (Edited by L’Harmattan).

He is now working with the “Painting Beings” (virtual creatures making Art) and the “Uchrons” (virtual creatures living with no linear time).

He is also an “Algoriste” (see ), and a member of the programme committee of EvoMusart : ()

Links for the paper :

Web site (with old works, soon the BlenderAI page)



Other important links for the Development Project Blender: BlenderAI

The new functions proposed are based on :

– Genetic Algorithm :

– Genetic Programming :

– Neural Network (Backpropagation) :

– Self Organizing Map :

– Principal Component Analysis :

– Bayesian Network :

– Fuzzy Logic :

Important works and research in this field :

– Karl Sims works ( )

– William Latham et Stephen Todd creations:

– Sensors Actuators Networks and other works from Michiel Van De Panne :

– Larry Gritz 's work (Pixar) on Genetic Programming for 3D Animation :

– Matthew Lewis works (MetaVolve and other projects) :

– Christian Jacob evolutionary works:

– Jeffrey Ventrella works, mainly Disney Meets Darwin :

– Penousal Machado (book and NevAr system):

– Michel Bret :

Some existing tools :

– Kandid ()

– Evolvotron :

– Texture Maker (Genetic Tool) : :

– ImageSpace :

– PicBreeder :

– ImageBreeder :

– Breed from Driessens and Verstappen :

– ArtMatic Pro :

– Neat Particles :

– xTNZ :

– Evolve (Python) :

etc.

The first commercial tool for 3D animation :

Endorphin from Natural Motion (based on Torsten Reil thesis) :

Recent Blender works in this field (Blender Conferences 2006 and 2007):

– 3D simulation from Albert Cardona (BC 2007)

– Friedrich Förstner's Neurosciences works (BC 2007)

– Swarms from Théo de Ridder (BC 2007)

– Skin reconstruction works from Albert Cardona (BC 2006)

– Sensors Networks from Théo de Ridder (BC 2006)

Some libraries written in Python :

– PyGP : Genetic Programming :

– PyGene : Genetic Algorithm :

– other GP system:

– other GA system :

– Principal Components Analysis :

– Bayesian Network :

– Fuzzy Locic :

– Neural Network :

– Self Organizing Map :

Some libraries in C, C++ :

– EO (GA):

– Open Beagle (GA, GP):

– FANN (Neural Network) :

– FFLL (Fuzzy Logic) :

– Dlib : (Bayesian Network, Machine Learning, ...) :

– SciLab (PCA, etc.) :

Bibliography, Books :

– The Art of Artificial Evolution : A Handbook on Evolutionary Art and Music . Springer Berlin Heidelberg. 2007

– Evolutionary Art and Computers, W Latham, S Todd, 1992, Academic Press

– MetaCreation. Art and Artificial Life. Mitchell Whitelaw. MIT Press.

– Many Books from Peter J.Bentley : (Digital Biology, ...)

– Emergence de Nouvelles Esthétiques du Mouvement. Alain Lioret. L'Harmattan. 2004. France

(with the GaBuZoR System, including some experimental modules before developping BlenderAI).

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download