Table 8 Zero-shot recognition results (± standard deviation) of the SAEVRT method compared with those of three variants (\(\%\)).
From: Enhancing zero-shot scene recognition through semantic autoencoders and visual relation transfer
Method | Scene15 | MIT67 | UCM21 | NWPU45 |
|---|---|---|---|---|
SAEVRT-\(\lambda _1\) | 56.55 ± 8.24 | 63.91 ± 7.55 | 52.09 ± 10.68 | 47.08 ± 5.96 |
SAEVRT-\(\lambda _2\) | 61.73 ± 10.53 | 66.45 ± 10.39 | 57.62 ± 11.81 | 51.75 ± 6.78 |
SAEVRT-\(\lambda _3\) | 59.46 ± 9.76 | 65.76 ± 10.41 | 56.89 ± 13.24 | 50.48 ± 6.39 |
SAEVRT | 63.77 ± 9.71 | 67.75 ± 10.35 | 58.68 ± 10.04 | 53.26 ± 6.11 |
