Archaeoinformatics - Data Science

BA/MA: Pattern Similarity Search on Neolithic Ceramic Sherds

Contact: Christian Beth M.Sc.

A routine task within the field of Archaeology involves the analysis of ceramic sherds from various dig sites in order to extract useful information, such as the time period or cultural background of the artisans that fashioned the ceramic. A common step in this analysis is to retrieve other collected ceramic sherds that depict similar patterns and shapes as decorations on the sherds. The goal of this thesis will be to effectively and efficiently retrieve sherds with similar patterns from a database in order to assist the archaelogogist in her search in order to avoid manual browsing and retrieval in the database.

There are multiple sub-problems which can - but do not have to - be addressed in the same thesis:

The first sub-problem is to provide query support in a database that enables convenient retrieval of sherds depending on common queries without the need for overly complex SQL-queries. This could encompass stratifying the DB schema for easier use or adding key-words to the query language to enable more convenient query formulation.

The second sub-problem involves the automatic pattern extraction from image with CNNs to enable the search for similar patterns in the database with the help of image features.