transfer learning note

The network before the last fully connected layer of Inception-v3 is called the bottleneck layer (bottleneck).
We first input a picture to the bottleneck layer, and the bottleneck layer outputs a feature vector. We use this feature vector to train the new fully connected layer.

(Looking for the concept wkl for a long time...)

https://blog.csdn.net/HUXINY/article/details/86658368

 

Guess you like

Origin blog.csdn.net/qq_41427834/article/details/105578310