Research Article
[Retracted] Identifying Animals in Camera Trap Images via Neural Architecture Search
Table 3
Accuracy comparison based on MCTI.
| Class | NASRT (ours) | Resnet-18 [46] | DenseNet [47] | MobileNet-v2 [48] | SGAS [44] | SETN [26] | PDARTS [24] |
| Agouti | 97.60 | 95.21 | 96.41 | 95.81 | 96.41 | 89.82 | 92.81 | Bird | 98.61 | 97.92 | 97.92 | 99.31 | 100.00 | 98.61 | 100.00 | Coiban agouti | 100.00 | 100.00 | 100.00 | 100.00 | 99.61 | 98.06 | 100.00 | Collared peccary | 98.25 | 96.49 | 94.15 | 92.98 | 97.66 | 92.40 | 98.83 | Opossum | 98.15 | 94.44 | 96.91 | 96.30 | 98.15 | 93.21 | 97.53 | European hare | 97.48 | 89.92 | 96.64 | 96.64 | 96.64 | 99.16 | 94.96 | Great tinamou | 99.56 | 99.11 | 98.22 | 99.56 | 99.56 | 93.33 | 97.78 | Mouflon | 99.77 | 99.30 | 99.30 | 99.07 | 99.30 | 97.91 | 98.84 | Ocelot | 89.47 | 82.11 | 85.26 | 75.79 | 82.11 | 68.42 | 74.74 | Paca | 100.00 | 92.09 | 96.74 | 95.35 | 95.81 | 90.70 | 96.28 | Red brocket deer | 96.59 | 93.75 | 96.02 | 93.18 | 91.48 | 93.18 | 92.61 | Red deer | 100.00 | 100.00 | 99.80 | 100.00 | 100.00 | 99.61 | 100.00 | Red fox | 95.60 | 93.41 | 92.31 | 91.21 | 93.41 | 79.12 | 93.41 | Red squirrel | 97.35 | 93.81 | 94.69 | 93.81 | 94.69 | 88.50 | 87.61 | Roe deer | 98.72 | 98.29 | 98.29 | 98.29 | 98.29 | 96.58 | 98.29 | Spiny rat | 99.24 | 100.00 | 96.21 | 91.67 | 97.73 | 84.85 | 97.73 | White-nosed coati | 99.57 | 99.15 | 98.72 | 97.02 | 97.87 | 95.32 | 97.45 | White-tailed deer | 99.75 | 98.99 | 97.98 | 98.24 | 99.50 | 95.97 | 99.75 | Wild boar | 99.71 | 98.82 | 99.41 | 99.71 | 99.41 | 99.41 | 99.71 | Wood mouse | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | Average | 98.27 | 96.14 | 96.75 | 95.70 | 96.88 | 92.71 | 95.92 |
|
|