Optimization and implementation of bio-inspired feature extraction frameworks for visual object recognition
2016
Industry has growing needs for so-called \intelligent systems", capable of not only acquire
data, but also to analyse it and to make decisions accordingly. Such systems are
particularly useful for video-surveillance, in which case alarms must be raised in case of
an intrusion. For cost saving and power consumption reasons, it is better to perform that
process as close to the sensor as possible. To address that issue, a promising approach is
to use bio-inspired frameworks, which consist in applying computational biology models
to industrial applications. The work carried out during that thesis consisted in selecting
bio-inspired feature extraction frameworks, and to optimize them with the aim to
implement them on a dedicated hardware platform, for computer vision applications.
First, we propose a generic algorithm, which may be used in several use case scenarios,
having an acceptable complexity and a low memory print. Then, we proposed optimizations
for a more global framework, based on precision degradation in computations,
hence easing up its implementation on embedded systems. Results suggest that while
the framework we developed may not be as accurate as the state of the art, it is more
generic. Furthermore, the optimizations we proposed for the more complex framework
are fully compatible with other optimizations from the literature, and provide encouraging
perspective for future developments. Finally, both contributions have a scope that
goes beyond the sole frameworks that we studied, and may be used in other, more widely
used frameworks as well.
Keywords:
- Field-programmable gate array
- Theoretical computer science
- Computation
- Artificial neural network
- Use case
- Intrusion
- Genetic algorithm
- Intelligent decision support system
- Feature extraction
- Machine learning
- Artificial intelligence
- Computer science
- Encoding (memory)
- Distributed computing
- Cognitive neuroscience of visual object recognition
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
105
References
0
Citations
NaN
KQI