Closed xiximelon closed 3 months ago
以下调用DCN部分的代码，fea = (self.dconv_1(fea, offset1))中self.dconv_1输入的是特征和偏移量，但是fea特征是连接了support帧和参考帧的特征，是否应该只输入supp的特征进行align? fea = torch.cat([ref, supp], dim=1) fea =self.cr(fea) offset1 = self.off2d_1(fea)\ fea = (self.dconv_1(fea, offset1))
以下调用DCN部分的代码，fea = (self.dconv_1(fea, offset1))中self.dconv_1输入的是特征和偏移量，但是fea特征是连接了support帧和参考帧的特征，是否应该只输入supp的特征进行align? fea = torch.cat([ref, supp], dim=1) fea =self.cr(fea) offset1 = self.off2d_1(fea)\ fea = (self.dconv_1(fea, offset1))