Table 4 Comparison of our proposed approach with state of the art algorithms (testing accuracy range), on different available datasets.
From: Gun identification from gunshot audios for secure public places using transformer learning
Model | Our dataset (%) | TREC (%) | Urban (%) |
|---|---|---|---|
Resnet50 (Baseline: raw audio) | 76.0–78.0 | 71.0–73.0 | 70.0–73.4 |
Capsule network37 | 82.2–83.6 | 83.1–84.3 | 80.1–81.9 |
DNN ensemble40 | 83.0–84.5 | 82.2–83.4 | 84.0–85.0 |
Zero-shot federated learning41 | 83.5–86.0 | 82.9–24.8 | 83.0–85.0 |
Resnet50+MFCC+MelSpectogram | 84.0–87.0 | 83.0–84.5 | 82.9–85.0 |
(PA) VT+MFCC+MelSpectogram | 89.0–90.0 | 88.0–89.5 | 87.4–89.0 |
