Date post: | 31-Mar-2015 |
Category: |
Documents |
Upload: | rudy-oswell |
View: | 214 times |
Download: | 0 times |
1
Neuromath kickoff meeting
2009-01-23
É. Tlapale1, G.S. Masson2, P. Kornprobst1
1 : Équipe Odyssée (INRIA, ENS, INPC)2 : Équipe DyVA (INCM/CNRS)
2
Current work• We design a bio-inspired model of motion integration.
– Functional model (V1/MT)– Feedbacks (MT => V1)– Form modulation (V2)
• Why ?– To reproduce motion perception– And its dynamic
3
Influence of form
Chopstick illusions
4
Influence of form
Chopstick illusions
5
Influence de la forme
Jonctions intrinsèques vs extrinsèques
6
Multi-layer feedback model
Input : grey-level sequence
7
Multi-layer feedback model
Local motion detectors: correlation or energy models
8
Multi-layer feedback model
9
Multi-layer feedback model
Lateral inhibition => winner-take-all
10
Un modèle multi-couche rétroactif
Feedback (Bayerl & Neumann, 2004)
11
Multi-layer feedback model
Anisotropic diffusion
12
Form-modulated diffusion
We only use luminosity for now
13
Reproducing the perception
Wallach (1935)
14
Wallace, Stone et Masson (2005)
Reproducing the dynamic
Eye-movement like read-out
15
Reproducing the neurobiology
Huang, Albright et Stoner (2007)
Notion of objects
16
Future work• Reformulate our model in a neural field formalism
• And using a multi-layer variational formalism
• Propose our form-motion interaction as a synchronisation mechanism
17
Journal clubs• Biology: Barthelemy et al. (2008)
• Models: Cesmely & Wang (2000)
• Technologies: Python, GPGPU