Rick Bassett's Coin Grading Dissertation Website - Pace University New York / Connecticut

Automated Coin Grader - Dissertation Website

By: Richard Bassett - 2003 DPS Student, Pace University

The Long Road....The dissertation process started for me in Sept 2001 when we had to come up with topics. Five ideas later and the actual writing started in May 2002, which should conclude with the defense sometime in April 2003 if everything goes according to plan. I am writing about 2 - 4 draft pages per week at this point and expect the finished dissertation product to be on the order of 100 - 130 pages when completed.

Revised Problem Statement (10/02): The focus of this study is to develop the systematic framework for a system that combines the computational power of a machine with skilled human interaction to provide a richer experience in the grading process of rare collectibles, such as rare coins.

Rationale / Justification: Rare coins are presently graded by human hand and eye inspection that often produces varied, inconsistent and sometimes dubious results. For instance one grader may assign a grade of Very Good to a particular rare coin and another grader may assign a grade of Fine to the same coin. On many rare coins a difference of a single grade can often mean thousands of dollars in difference in the value of the asset. Sometimes these discrepancies in grades are simply errors by the graders due to poor training, poor lighting, fatigue or misinformation. But many times the discrepancies can be attributed to dealers under-grading items so that they can purchase them for an amount that is cheaper than what they are worth or over-grading them so that they can sell the items for more than they are worth.

Initial Research: Our team of graduate students was able to develop an application last spring that was capable of performing fully automated technical grading at a consistent by rudimentary level. The software simply read the asset by scanning it into a digitized format and produced an automated technical assessment. Overall the results were encouraging but we discovered that by totally eliminating some important subjective human factors in the grading process such as: eye appeal, toning, aesthetic quality and the noting of damages that results lacked acceptance amongst the experts.

Second Round of Research: To overcome the objections encountered in the initial technical grading research, it was determined that human interaction needed to be included in the process of grading collectibles. Thus the systematic framework would now be extended to have to critical elements:

1. Fully automated grading (aka Technical Grading) - Upon the examination of the scanned collectible the fully automated grading process will return a technical grade in the range of 1 - 70.

2. Interactive human / machine grading (aka Hybrid Grading) - The output from technical grading process can be evaluated and changed and added to by a domain expert to produce a new grade that is a composite of the technical grade and the interactive grade. Click here to view the current human/machine interface (it is a Work in Process, requires Flash 6.0)

Important Project Links:

Motivation for this Project

PowerPoint Overview - July 27, 2002

PowerPoint Project Update - Sept 29, 2002

PowerPoint Progress Update - Nov 4, 2002

PowerPoint Progress Update - Dec 6, 2002

PowerPoint Progress Update - Feb 7, 2003

Technical Paper on Direction - March 8, 2003 (PDF Format)

PowerPoint Progress Update - March 14, 2003

PowerPoint Progress Update - April 25, 2003

PowerPoint Presentation - Student Research Day - May 9, 2003

PowerPoint Status - May 29, 2003