This document describes research on bio-inspired active vision systems. It discusses how biological vision differs from traditional computer vision in being active rather than passive. The researchers are developing active vision systems using an evolutionary robotics approach, involving neural networks and genetic algorithms. Previous related work is described, including obstacle avoidance by Mars rovers and koala robots. The document outlines plans to design an active vision system to recognize objects using a dataset of images under different conditions, and accelerate it with GPUs. Results showed the system learned to correctly classify objects over generations.

1. Bio-inspired Active Vision
Martin Peniak, Davide Marocco, Ron Babich and John Tran

2. Outline
Introduction
Classical vision vs. active vision
Show few examples, e.g. template matching, face detection etc.
Pros&Cons
Explain what is active perception (use current + davides slides)
Active Vision – many possibilities but we like GA+NN, why?
Based on neural network and genetic algorithm
What is neural network?
What is genetic algorithm?
Pros&Cons
Little computation required due to neural network architecture, fast
No external representation needed thanks to GA evolving nn weights
Generality, invariance, and application in multiple domains
Cons: need more research, so far applied in limited domains
Previous research by Floreano, Davide, me (ESA)
Show videos and pictures
Describe work done at NVIDIA
Future work: AXA grant
2

3. GPU Computing Lab

4. Traditional Computer Vision
4

5. Traditional Computer Vision
“Teaching a computer to classify objects has proved much harder than was originally anticipated”
Thomas Serre - Center for Biological and Computational Learning at MIT
Specific template or computational
representation is required to allow object
recognition
Must be flexible enough to account with all
kinds of variations
5

6. Biological Vision
“Researchers have been interested for years in trying to copy biological vision systems,
simply because they are so good” ~ David Hogg - computer vision expert at Leeds University, UK
Highly optimized over millions of years of
evolution, developing complex neural structures
to represent and process stimuli
Superiority of biological vision systems is only
partially understood
Hardware architecture and the style of
computation in nervous systems are
fundamentally different 6

7. Biological Vision
7

8. Seeing is a way of acting
8

9. Active Vision
Inspired by the vision systems of natural organisms that have
been evolving for millions of years
In contrast to standard computer vision systems, biological
organisms actively interact with the world in order to make sense
of it
Humans and also other animals do not look at a scene in fixed
steadiness. Instead, they actively explore interesting parts of the
scene by rapid saccadic movements
9

10. Creating Active Vision Systems
Evolutionary Robotics Approach
10

11. Evolutionary Robotics
New technique for the automatic creation of autonomous robots
Inspired by the Darwinian principle of selective reproduction of
the fittest
Views robots as autonomous artificial organisms that develop
their own skills in close interaction with the environment and
without human intervention
Drawing heavily on biology and ethology, it uses the tools of
neural networks, genetic algorithms, dynamic systems, and
biomorphic engineering
11

12. Genetic Algorithms (GAs) are adaptive heuristic search
algorithm premised on the evolutionary ideas of natural
selection and genetic. The basic concept of GAs is
... designed to simulate processes in natural system
necessary for evolution.
Population
(Chromosomes)
...
...
Genetic Evaluation
operators (Fitness)
Artificial neural networks (ANNs) are very powerful brain-inspired Selection
(Mating Pool)
computational models, which have been used in many different
12
areas such as engineering, medicine, finance, and many others.

13. Related Research
Mars Rover obstacle avoidance (Peniak et al.)
13

14. Related Research
Koala robot obstacle avoidance (Marocco et al.)
14

15. Related Research
Autonomous driving car (Floreano et al.)
15

16. Related Research
Object recognition (Floreano et al.)
16

17. Going Further
Designing active vision system for real-world object recognition
17

18. Task
Design active vision system that can learn to recognize the
following objects
18

19. Method
Evolution of the active vision system for real-world object recognition
training the system in a parallel manner on multiple objects viewed from many different angles and under different lighting conditions
Amsterdam Library of Object Images (ALOI)
provides a color image collection of one-thousand small objects
recorded for scientific purposes
systematically varied viewing angle, illumination angle, and illumination color
Active Vision Training
trained on a set of objects from the ALOI library
each genotype is evaluated during multiple trials with different randomly rotated objects and under varying lighting conditions
evolutionary pressure provided by a fitness function that evaluates overall success or failure of the object classification
trained on increasingly larger number of objects
Active Vision Testing
robustness and resiliency of recognition of the dataset
generalization to previously unseen instances of the learned objects
19

20. Experimental Setup
Recurrent Neural Network
Inputs: 8x8 neurons for retina, 2 neurons for proprioception (x,y pos)
No hidden neurons
Outputs: 5 object recognition neurons, 2 neurons to move retina (16px max)
Genetic Algorithm
Generations: 10000
Number of individuals: 100
Number of trials: 36+16 (object rotations + varying lighting conditions)
Mutation probability: 10%
Reproduction: best 20% of individuals create new population
Elitism used (best individual is preserved)
20

21. Experimental Setup
Each individual (neural network) could freely move the retina and
read the input from the source image (128x128) for 20 steps
At each step, neural network controlled the behavior of the
system (retina position) and provide recognition output
The recognition output neuron with the highest activation was
considered the network’s guess about what the object was
Fitness function = number of correct answers / number of total steps
21

22. GPU Accelerating GA and ANN
GPUs were used to accelerate:
Evolutionary process – parallel execution of trials
Neural Network – parallel calculation of neural activities
22

23. Results
Fitness can not reach 1.0 since it takes few time-steps to recognize an object
All objects are correctly classified at the end of the each test
0.9
0.8
0.7
0.6
0.5
fitness
0.4
0.3
0.2
0.1
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
generations
best fitness average fitness
23

24. Evolved Behavior
24

25. Future Work
25

26. "Imagination is the highest form of research"
Albert Einstein
Thank you!
26