Previous article implemented Gan handwriting recognition in Keras. At that time, I was curious about how Gan's learning process was, so I made a video.
I wanted to make a comparison this time, so I did it with Gan and CCan.
If you look at the video, you can see in the video how Gan captures the characteristics of handwritten characters. If you compare the two videos, you can see that the learning situation is completely different between GGan and Gan, and that CGan has the characteristics of the convolution layer firmly. The video is uploaded on Youtube.
GAN : https://www.youtube.com/watch?v=ORVZVZqYYqU CGAN : https://www.youtube.com/watch?v=ByicWghi-iw
The only difference between the two is the type of layer used in the network. The network is constructed with Gan as a fully connected layer and CGan as a convolution layer.
In the video, you can clearly see the difference in learning between layers.
I found it interesting to take a look at the learning process of the network!
https://www.youtube.com/watch?v=ORVZVZqYYqU
https://www.youtube.com/watch?v=ByicWghi-iw
Recommended Posts