ai_project(group 06)
TRANSCRIPT
AI Based System for Recognizing
Handwritten (Graffiti) Numerals
CMIS 4114 – Artificial Intelligence
Mini Project – Group No. 06
Group Members
• 112212 – R. A. N. C. Ranasinghe• 112217 – W. M. B. Nayanadarshani
Problem Definition• Digital Character Recognition is one of the low-end
applications of the AI.• Since, the handwritings are different from person-to-
person, simple scanning or pixel based identification is not very effective.
• Therefore, such systems can be designed / implemented by training an AI agent to identify graffiti numerals.
• A graffiti inserted through a mouse or touch device has to be identified without depending on the persons’ unique writing characteristics.
Technologies
• Microsoft Visual Studio IDE• Visual Basic
Neural Networks
• A Neural Network, consisted of 03 neuron layers is implemented to achieve the goal.
• In the Network, layers are connected following a Learning Vector Quantization (LVQ) algorithm and Self-Organizing Maps operate hand-to-hand to analyze, identify and to deliver the final result.
AI Algorithms• Learning Vector Quantization (LVQ)
This is a supervised learning system where the training process is defined by the user. The system does not have a topological structure.
Demonstration
Pros & Cons Pros Cons
User-friendly GUI The System struggles and sometimes give false outputs to the similar digits (1 and 7, 3 and 5, 8 and 9).
User-definable training process
The System fails to self-learn the patterns.
Relatively simple operation
No feedback database is maintained for tried out patterns.
Small, compact size. The necessity of intensive training
The input matrix is expandible
The mouse is not a flexible input device
Thank You. . !