Table 1 Results of random forest classifications from experiments with instrumented knife. This includes both detecting contact types (Top) and detecting if the knife is approaching contact (Bottom).
From: Safely and autonomously cutting meat with a collaborative robot arm
Data | Action | False positives | False negatives | Total occurrences | Error rate (%) |
|---|---|---|---|---|---|
Superficial, within cut type (SWT) | Slicing | 86 | 168 | 13,670 | 1.86 |
Trimming | 143 | 300 | 11,998 | 3.69 | |
Cubing | 236 | 328 | 18,779 | 3.00 | |
By replicate, within cut type (RWT) | Slicing | 30 | 304 | 8123 | 4.11 |
Trimming | 147 | 79 | 11,948 | 1.89 | |
Cubing | 826 | 779 | 16,769 | 9.57 | |
Superficial, across types (SAT) | All cuts | 802 | 1148 | 44,090 | 4.42 |
SWT approaching contact | Slicing | 0 | 88 | 13,438 | 0.007 |
Trimming | 2 | 91 | 12,135 | 0.008 | |
Cubing | 0 | 108 | 18,924 | 0.006 | |
RWT approaching contact | Slicing | 0 | 87 | 8123 | 0.011 |
Trimming | 0 | 17 | 12,135 | 0.001 | |
Cubing | 0 | 133 | 16,769 | 0.008 | |
SAT approaching contact | All cuts | 5 | 305 | 44,448 | 0.007 |
