ICPUTRD: Image Cloud Platform for Use in Tagging and Research on Decomposition

As submitted by the proposer:
The quantification of human decomposition is an important problem in forensic anthropology. The ability to conduct rigorous statistical analysis of large data collections available in, for example, the Anthropology Research Facility (ARF) at the University of Tennessee, has the potential to transform research and practice for this domain. The ARF has over a million photographs of more than 400 donor cadavers. Multiple images of each donor are taken daily documenting the decomposition process. Images in the ARF’s collection are, at present, very difficult to use in research or practice because of the of the collection’s size, the varying quality of the images, the lack of tools to find and retrieve relevant images, and the lack of forensics-relevant metadata associated with these images. We propose to develop a standard nomenclature of trait observations in layman’s terms for tagging images, to stitch the images showing the different parts of the body into a single image, and to register the stitched images taken at subsequent days for the same donor to create spatio-temporal models of the decomposition process. Raw, stitched, and registered images in conjunction with the derived forensics-relevant statistics will be used to create a searchable and expandable collaboration platform to make it easy to conduct replicable research on the decomposition of human remains. In particular, the platform would have the functionality to find images based on the abovementioned nomenclature or based on spatio-temporal properties, such as unusual volumetric or texture changes in specific areas. The platform will be both highly scalable and easily extensible with additional research tools or annotations. To evaluate and demonstrate the capabilities of the platform we will develop a computer-assisted approach to tag tens of thousands of images with very high accuracy using the developed forensics-specific nomenclature. This proposed work, if funded, would make complex image data stored in ARF and similar facilities understandable and accessible to a broad group of researchers and practitioners and would provide future decomposition research with a rigorous empirical basis.

Note: This project contains a research and/or development component, as defined in applicable law.

ca/ncf