Raw data of the error analysis of krill, pelagic shrimp, and gelatinous zooplankton on all three test sets.

Columns: image (name of image file), error_type (FP or FN); class (object class according to human annotation or detection); x1, y1, x2, y2 (coordinates of bounding box in pixels); score (confidence score of FP detection); test_set (white, red gain 1.5, red gain 5 test set); directory (folder directory of the image with information of non-maximum suppression IoU and confidence threshold that was applied); score_threshold (optimal score threshold as determined by F1-confidence curve); below_above_score_theshold (label indicating if a FP detection was below or above the optimal score threshold); id (unique ID for each FP or FN bounding box); width (width of the image in pixels); height (height of the image in pixels); sample (“yes” if included in random sample, “no” if not included in random sample for the analysis conducted to produce Fig 11 in the manuscript); contrast (“good”, “average”, “bad” contrast of an object with the background), overlap (“yes”, “no”, overlap with another object), visible_morphological_features (visible morphological features of an object); linked_id (id of a detection or annotation of the same object); cause (immediate cause resulting in a FP or FN label); cause_duplicate (cause leading to a duplicate detection); cause_misclassified (cause leading to a misclassification); class_2 (correct class of a misidentified object). (CSV)