本筆記主要紀錄image segmentation相關論文。
Fully Convolutional Networks for Sematic Segmentation (FCN)^1
14 Nov 2014
貢獻
第一個提出全卷積網路概念，利用卷積網路保留特徵位置資訊。
利用影像分類的既有模型當作編碼器(Encoder)，汲取不同層次特徵加以結合，預測分割圖。
實測AlexNet(39.8)、VGG16(56.0)、GoogLeNet(42.5)
Su Yan-Ru changed 5 years agoView mode Like BookmarkAre you sure to leave this team?
Once you delete your team, all team notes will be deleted and cannot be recovered. Please ensure you've exported or transfered these notes.
Enter team name before deleting it:
or
By clicking below, you agree to our terms of service.
Sign in with Wallet
New to HackMD? Sign up
