Skip to main content
. 2023 Nov 7;101:skad376. doi: 10.1093/jas/skad376

Table 2.

Validation results for sensitivity, precision, and F1-score in percentage for DL, using the 7:3 random split

BCS Sensitivity, % Precision, % F1-Score, %
Exact 0.5-unit DEV Exact 0.5-unit DEV Exact 0.5-unit DEV
1.5 3 (0-11) 88 (71-100) 25 (0-100) 40 (0-100) 3 (0-14) 39 (0-100)
2.0 55 (45-62) 99 (97-100) 63(60-69) 98 (96-99) 59 (51-63) 98 (97-99)
2.5 67 (63-76) 98 (96-100) 47 (43-52) 94 (93-97) 55 (52-57) 96 (95-97)
3.0 33 (21-46) 98 (96-99) 40 (35-45) 91 (88-97) 36 (27-43) 94 (93-97)
3.5 44 (32-60) 82 (77-93) 41 (32-51) 88 (82-93) 42 (32-48) 85 (80-92)
4.0 7 (0-29) 72 (61-83) 30 (0-100) 93 (71-100) 9 (0-34) 81 (73-91)
4.5 3 (0-20) 26 (0-55) 13 (0-100) 23 (0-100) 4 (0-25) 13 (0-57)
WAvg 48 (46-51) 94 (93-95) 47 (44-52) 91 (89-94) 46 (44-49) 91 (89-94)

The parenthesis represents the range among replicates. DL = DeepLearning, BCS = Body Condition Score, Exact = exact score, DEV = deviation, WAvg = weighted average by frequency.