Table 6 Effects of different strategies on the results
From: ANINet: a deep neural network for skull ancestry estimation
Convolutional network | Accuracy (%) | Para size | |||
|---|---|---|---|---|---|
Local depth images | AlexNet | 90.83 | − 7.21 | 45 MB | − 4 k |
SE-AlexNet (SE block) | 92.83 | − 5.21 | 45 MB | – | |
Wide-AlexNet(Wide) | 96.58 | − 1.46 | 45 MB | -4 k | |
ANINet(SE-Wide) | 98.04 | – | 45 MB | – | |
Global depth image | AlexNet | 95.04 | − 3.17 | 45 MB | − 4 k |
SE-AlexNet(SE block) | 95.93 | − 2.28 | 45 MB | – | |
Wide-AlexNet(Wide) | 97.47 | − 0.74 | 45 MB | − 4 k | |
ANINet(SE-Wide) | 98.21 | – | 45 MB | – | |
Local + global depth image | AlexNet | 95.56 | − 3.47 | 45 MB | − 4 k |
SE-AlexNet(SE block) | 97.91 | − 1.12 | 45 MB | – | |
Wide-AlexNet (Wide) | 98.43 | − 0.60 | 45 MB | − 4 k | |
ANINet (SE-Wide) | 99.03 | – | 45 MB | – | |