Semiautomatic Expansion of the Dataset
Our tkinter GUI allows for direct comparison of our existing seal database with new probed individuals, supported by similarity scores of the individuals in the gallery to the current probe (Figure 4 ). These similarity scores are based on facial biometrics of each individual, and they are determined after each image within a seal individual folder is compared to each image found in a gallery seal individual folder. On average, the similarity scores for individuals that are a match with the probe individual are approximately 0.65 with a SD of 0.1 (range 0.55-0.85). Similarity scores with individuals without a match range in similarity scores from 0.2-0.5.