Table 4 Comparison of different infrared small target detection methods on the NUDT-SIRST dataset.
From: Deep asymmetric extraction and aggregation for infrared small target detection
| Â | Method | \(IoU(\times 10^{-2})\) | \(P_{d}(\times 10^{-2})\) | \(F_{a}(\times 10^{-5})\) | Time (s) | FLOPs (G) | Params (M) |
|---|---|---|---|---|---|---|---|
Model-Driven | Top-Hat10 | 20.83 | 78.41 | 107.54 | 0.0288 | - | – |
Max-Median11 | 4.56 | 73.02 | 990.799 | 0.0176 | - | – | |
WSLCM29 | 6.82 | 88.47 | 2098.72 | 14.1103 | – | – | |
TLCM30 | 3.72 | 32.01 | 1162.42 | 2.9738 | – | – | |
IPI15 | 23.41 | 79.47 | 53.32 | 0.5691 | – | – | |
NRAM31 | 7.42 | 58.31 | 10.14 | 4.2193 | – | – | |
RIPT16 | 30.41 | 92.06 | 185.31 | 1.0926 | – | – | |
PSTNN32 | 14.87 | 66.98 | 29.06 | 0.7275 | – | – | |
MSLSTIPT33 | 12.87 | 62.86 | 24.56 | 0.083 | – | – | |
Data-Driven | U-Net20 | 75.58 | 96.72 | 24.63 | 0.069 | 0.31 | 1.6 |
ACM5 | 70.56 | 97.24 | 25.36 | 0.078 | 0.30 | 1.6 | |
ALC12 | 73.79 | 97.86 | 21.09 | 0.072 | 0.41 | 1.44 | |
DNANet22 | 88.91 | 99.25 | 2.34 | 0.064 | 10.91 | 18.7 | |
ISTDU-Net13 | 86.47 | 97.98 | 3.71 | 0.072 | 6.08 | 11.3 | |
DAEA(S=3) | 75.37 | 97.52 | 20.61 | 0.064 | 1.96 | 1.2 | |
DAEA(S=4) | 77.94 | 98.20 | 14.63 | 0.066 | 2.43 | 1.6 | |
DAEA(S=5) | 81.32 | 98.48 | 7.23 | 0.072 | 2.90 | 2.0 | |
DAEA(S=6) | 80.26 | 98.10 | 12.45 | 0.076 | 3.35 | 2.4 |
