Table 8 Total products considered (Total), accepted products (Accept), rejected products (Reject), test duration in minutes (M), true positives (TP), true negatives (TN), false positives of the deterministic procedure (\(FP_D\)), false positives of the CNN model ( FP), false negatives of the deterministic procedure ( \(FN_D\)), false negatives of the CNN model (FN), false omission rate (FOR) and false discovery rate (FDR) for each phase (P) of the real scenario test with the final pre-trained DenseNet161. Percentages have been computed from total production.
From: Deep learning for the quality control of thermoforming food packages
P | Total | Accept | Reject | M | TP | TN | \(FP_D\) | FP | \(FN_D\) | FN | FOR | FDR | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1 | 3098 | 3066 | 32 | 1.03% | 60 | 12 | 3065 | 13 | 0.42% | 7 | 0.23% | 1 | 0.03% | 0 | 0% | 0 | 0.90 |
2 | 2801 | 2755 | 46 | 1.64% | 30 | 24 | 2755 | 16 | 0.57% | 6 | 0.21% | 0 | 0% | 0 | 0% | 0 | 0.48 |
3 | 2880 | 2858 | 22 | 0.76% | 75 | 21 | 2857 | 1 | 0.03% | 1 | 0.03% | \(<0.001\) | 0.045 | ||||
