Transcript Analog Neural Network Hardware For Colour Classification

Analog Neural Network Hardware For
Colour Classification
Abdullah VURAL
September 07, 2006
Prof. Dr.-Ing. Andreas König
Analog Neural Network Hardware
For Colour Classification






An Overview to ANN
Objective and Our Case
Different Design Techniques
Neurokit Workflow and Network
Configuration
Implementation Phase
Assessment
Overview

Artificial Neural Networks(ANN)




Signal Processing
Control Systems(intelligent control sys.)
Industrial Automation(robot)
Adaptive Channel Equalization
Analog Neural Network Hardware
For Colour Classification






An Overview to ANN
Objective and Our Case
Different Design Techniques
Neurokit Workflow and Network
Configuration
Implementation Phase
Assessment
Objective : Pill Detection
Silimann
Trainer
Software
True Colour
Sensor
Evaluation Board
Our Case

Is PILL present or absent?
Analog Neural Network Hardware
For Colour Classification






An Overview to ANN
Objective and Our Case
Different Design Techniques
Neurokit Workflow and Network
Configuration
Implementation Phase
Assessment
Techniques(1)

Conventional algorithm/model based
design flow
Techniques(2)

Neural network based design flow

Is PILL present or absent?
Analog Neural Network Hardware
For Colour Classification






An Overview to ANN
Objective and Our Case
Different Design Techniques
Neurokit Workflow and Network
Configuration
Implementation Phase
Assessment
Neurokit Workflow
Sample Patterns
Perform New
Training
Silimann LSI
Run Task
Manager
SILIMANN TRAINER Software
Board Test
Run Pattern Test
Network Configuration
INPUTS
OUTPUTS
Input neurons Hidden neurons Output neurons
Analog Neural Network Hardware
For Colour Classification






An Overview to ANN
Objective and Our Case
Different Design Techniques
Neurokit Workflow and Network
Configuration
Implementation Phase
Assessment
Implementation Phase(Pictures)
Implementation Phase(Insufficiency)

Light Effect

Insufficient Tools

Sampling Errors
Implementation Phase




(What we learned)
Analog Neural Network Hardware(LSI)
ANN Design Flow
Evalution Board and Accessories
Technical Real World Imlementations

(e.g. Signal Processing, Industrial
Automation, Control Systems, Channel
Equalization)
Analog Neural Network Hardware
For Colour Classification






An Overview to ANN
Objective and Our Case
Different Design Techniques
Neurokit Workflow and Network
Configuration
Implementation Phase
Assessment
Assessment