Automatic Metadata Generation and Verification for AI-Ready Fish Image Collections

This poster covers recent efforts to develop automatic metadata generation and verification methods that are applied to images of fish specimens found in zoological museum collections. These methods extract, and in some cases verify, metadata important for subsequent image analysis associated with the individual specimens. Acquiring and verifying these metadata is an important step in making digital specimen collections AI-ready, i.e. copacetic for downstream machining learning analysis. Previous work by our group in this area is reviewed, including image quality assessment, object detection and segmentation, scale determination, quantitative feature calculation, label data verification and shape feature extraction.