Table 5 Accuracy, loss results and time taken to train the MTL models.
From: Real-time facial recognition via multitask learning on raspberry Pi
Model | Person | Age | Ethnicity | Â | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Training accuracy | Training loss | Test accuracy | Test loss | Training accuracy | Training loss | Test accuracy | Test loss | Training accuracy | Training loss | Test accuracy | Test loss | Time Spent (Seconds) | |
MTL-InceptionV3 | 0.9460 | 0.1686 | 0.9333 | 0.2120 | 0.9757 | 0.0783 | 0.9556 | 0.1206 | 0.9712 | 0.0932 | 0.9753 | 0.1209 | 4098.19 |
MTL-MobileNet | 0.9905 | 0.0686 | 0.9901 | 0.0465 | 0.9884 | 0.0411 | 0.9926 | 0.0200 | 0.9915 | 0.0369 | 0.9951 | 0.0212 | 2867.94 |
MTL-MobileNetV2 | 0.9852 | 0.0669 | 0.9827 | 0.0681 | 0.9873 | 0.0411 | 0.9728 | 0.0664 | 0.9873 | 0.0456 | 0.9901 | 0.0337 | 3712.28 |
