①のネットワークの学習時に、Global average pooling(KernelShape、Strideプロパティが入力画像の縦横サイズであるとしたAveragePooling)を行った後Affineを行うようなネットワーク構成を用いることで、画像サイズを変更してもパラメータサイズが変化しないようにする方法もあります。
watch
unread,
Sep 12, 2018, 1:02:29 AM9/12/18
Reply to author
Sign in to reply to author
Forward
Sign in to forward
Delete
You do not have permission to delete messages in this group
Copy link
Report message
Show original message
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to Neural Network Console Users (JP)
小林様
日頃よりお世話になっております.お返事いただき誠にありがとうございます.
>①のネットワークの学習時に、Global average pooling(KernelShape、Strideプロパティが入力画像の縦横サイズであるとしたAveragePooling)を行った後Affineを行うようなネットワーク構成を用いることで、画像サイズを変更してもパラメータサイズが変化しないようにする方法もあります。
「①のネットワークの学習時に」とは
a.①のネットワークを学習するときに
b.①のネットワークを利用して②を学習するときに
のどちらのことを指していますでしょうか?
以上,よろしくお願いします.
小林由幸
unread,
Sep 19, 2018, 9:19:15 PM9/19/18
Reply to author
Sign in to reply to author
Forward
Sign in to forward
Delete
You do not have permission to delete messages in this group
Copy link
Report message
Show original message
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to Neural Network Console Users (JP)
転移学習を目的とする場合、a、bどちらでも期待通りの動作が得られるかと思います。 ①のモデルを用いて何度も転移学習を行いたい場合は、aの時点でGlobal Average Poolingを用いる構成にしておいた方が転移学習の度にネットワークを編集する必要がなくなり、より良いと言えます。 また、aの時点でGlobal Average Poolingを用いておくことは、転移学習以外にサイズ違いの画像に認識を適用したい場合にもネットワーク編集の必要がなく、おすすめです。