Table 4.
Confusion Matrix of the best MVFCNN approach.
| Actual Class Labels | ||||||
|---|---|---|---|---|---|---|
| Martensite | Temp. Martensite | Bainite | Pearlite | Class Precision | ||
| Predicted Class Labels | Martensite | 1190 94.97% |
0 0.00% |
11 3.19% |
0 0.00% |
99.08% |
| Temp. Martensite | 24 1.92% |
268 97.81% |
0 0.00% |
0 0.00% |
91.78% | |
| Bainite | 39 3.11% |
6 2.19% |
325 94.20% |
16 4.80% |
84.19% | |
| Pearlite | 0 0.00% |
0 0.00% |
9 2.61% |
317 95.20% |
97.23% | |
| Class Recall | 94.97% | 97.81% | 94.20% | 95.19% | Total Accuracy 95.23% | |
The matrix shows the number of samples for each class predicted by the system. Due to an unbalanced multi-class problem, percentage numbers for each class show normalized recall rates. Note: #48 not-segmented objects have not been considered. Overall accuracy is 93.94%.