Open yusukekyokawa opened 3 years ago
https://arxiv.org/abs/1703.06211
本質的に、CNN はその構造上、幾何的な 1 変換処理に制限されている その欠点を克服するために、 deformable convolution と deformable RoI pooling を提案する 物体検出と semantic segmentation のタスクで結果を検証する
書誌情報
論文リンク
https://arxiv.org/abs/1703.06211
著者/所属機関
conf/journal
year
どんな論文か?
本質的に、CNN はその構造上、幾何的な 1 変換処理に制限されている その欠点を克服するために、 deformable convolution と deformable RoI pooling を提案する 物体検出と semantic segmentation のタスクで結果を検証する
新規性
手法
結果
コメント