Skip to main content

Table 1 Results of practical test for model (AI-PIC20191024) using samples U1422C, 1H-1, 68–70 cm (Japan Sea) and DCR1PC, #39 (Southern Ocean)

From: Automated collection of single species of microfossils using a deep learning–micromanipulator system

Item

U1422C; 1H-1, 68–70 cm

DCR-1PC, #39

(a) Confidence level

0.6

0.8

0.9

(b) Total number of images

2,472

11,416

9,815

(c) Actinomma boreale

45/48 (94%)

59/59 (100%)

41/41 (100%)

(d) Cycladophora davisiana

165/176 (94%)

103/149 (70%)

54/59 (92%)

(e) Uncategorized image

267 (11%)

4,335 (38%)

5,768 (59%)

  1. (a) confidence level (0.00–1.00) meaning the certainty of correct classification, (b) total number of images used in classification through the scan, (c–d) accuracies (number of correct classification/number of categorized images) for Cycladophora davisiana and Actinmma boreale, and (e) number of uncategorized images with percent of total