原创 ResNeSt: Split-Attention Networks

Abstract: 由於RexNet簡單和模塊化的結構,許多計算機視覺下游任務仍在使用。我們提出了Split-Attention block能夠使注意力across feature-map groups。並且不會引入額外的及計算消

原创 Context Prior for Scene Segmentation(CVPR2020)上下文優先的語義分割

Abstract: 研究上下文語義的依賴性爲得到精準的分割結果,但大多數研究都是區分不同types的語義依賴性,這會損失scene understanding。 本文設計了Context Prior Network(CPNet)

原创 mobilenet v1筆記

**Mobilenet v1** Abstract (1)use depth-wise separable convolutions to build light weight deep

原创 ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation筆記

ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation (a)是intial block。Input爲512x512。Num_fi

原创 語義分割的門控形狀CNN Gated-SCNN: Gated Shape CNNs for Semantic Segmentation

語義分割的門控形狀CNN Gated-SCNN: Gated Shape CNNs for Semantic Segmentation Abstract: (1)提出一種新的用於分割two-stream cnn 結構:shape

原创 deeplabv3+ 筆記

Deeplab v3+ Abstract (1)Deeplab v3+ adds a decoder modeule to refine the segmentation result along object boundarie