Progress Report

October 11^th, 2005

Overview

• MIAS-IRC Abstract
• ISBI submission outline
• Thesis structure
• Errors in sensitivity plots
• Refined and final figures from recent work

MIAS-IRC Abstract

• Template in finally in tact
• MIAS-IRC abstract compiled as a PDF and sent by E-mail
• Need to confirm or at least agree on prospective title, name of authors by the 14^th
• The remainder of the document is last year's content (without the images) so it should be ignored for the time being

ISBI Submission Plan

• Has not been reviewed since last changes had been made
• ISBI Structure Plan (PDF)

Thesis Structural Outline

• Has not been reviewed since last changes had been made
• Could benefit from further extension and embedment of experiments as placeholders
• Thesis Structure Plan (PDF)

Sensitivity Errors

• Previously discussed the possibility of watching the scatter of measures in the sensitivity plot
• Derive the errors over the sensitivity based on the scatter
• Issue: need error bars to derive sensitivity
• Without a collection of measures, we haven't any error bars
• Must stick to analytical solution (error propagation)

Sensitivity and Error Calculation

• Composition of errors was discussed in last meeting/s
• Errors have not been computed properly yet
• Errors in Sensitivity (PDF)

Generalisation and Error Bars

Generalisation ability plot

Specificity and Error Bars

Specificity plot

Sensitivity of Generalisation with Error Bars

Sensitivity of Generalisation ability (ignore errors)

Sensitivity of Specificity with Error Bars

Sensitivity of Specificity ability (ignore errors)

Sensitivity of Overlap

Sensitivity of overlap-based evaluation

Present and Future

• MIAS-IRC
  • Authors and title by the end of the week
  • 2-page abstract due by October 24^th
• ISBI submission - November deadline (moved to 15^th)
• Thesis structure - discussion and continued expansion
• Involvement in model building

Present and Future - Ctd.

• IMED implementation in C++
• ISBI and MIAS-IRC submission - drafts to be sent around
• Longer term goals:
  • Journal paper (possibly TMI)
  • Thesis chapters
  • Publication on face perturbation and evaluation - targetted at a general machine vision conference with broader scope