--- tags: Personal meeting title: Personal meeting 2021/03/02, 寒假報告 date: '02, March, 2021' --- ## Paper survey ### ResNeSt: Split-Attention Networks(2020/04/19,amazon & U.C. Davis) link: https://arxiv.org/pdf/2004.08955.pdf ### 概念 以SE與SK net的概念改進原有的ResNext 主要貢獻: 1.提出Split-Attention block 2.提出ResNest ### ResNest block  ### Split Attention in ResNest  ### Concurrent Spatial and Channel ‘Squeeze & Excitation’ in Fully Convolutional Networks link:https://arxiv.org/pdf/1803.02579.pdf 依照SE-Net提出三種不同的衍伸變體  ## linear attention 近期有部分研究開始著墨於直接用attention訓練(取代CNN) 但感覺還是在發展階段,因為attention本身計算量不低,目前主流應用還是需與RNN、CNN結合 ## 實驗部分 YOLOV3在detection應該是可以,但在VOC2007中loss可以降到0.8左右,但葉片的只會到12左右,覺得可能要增加數據並訂出一定的準則來框葉片會比較好
×
Sign in
Email
Password
Forgot password
or
By clicking below, you agree to our
terms of service
.
Sign in via Facebook
Sign in via Twitter
Sign in via GitHub
Sign in via Dropbox
Sign in with Wallet
Wallet (
)
Connect another wallet
New to HackMD?
Sign up