December 2018
Intermediate to advanced
764 pages
18h 18m
English
The DeepNet model is used for learning the embedding of faces for face verification tasks such as FaceNet. This improves on the method of FaceNet discussed in the previous section. It takes multiple crops of the same face and passes through several encoders to get a better embedding. This has achieved a better accuracy than FaceNet but takes more time for processing. The face crops are made in the same regions and passed through its respective encoders. Then all the layers are concatenated for training against the triplet loss.
Read now
Unlock full access