Xinyi Ying

Thanks for this comment. We have conducted several experiments to investigate the influence of deformation in temporal dimension. Specifically, we relaxed the temporal deformation constraint to perform deformations in all...

训练崩了吧，重新从没崩的那个epoch继续训，可以把学习率调小一点

Hi Shujun, We mainly use "Microsoft Visio" and "Microsoft PowerPoint" to make Fig. 1 in my paper. To be specific, we use "Microsoft Visio" to make the arrows and texts....

Thanks for this comment. We have compared our D3Dnet with both EDVR and DUF-VSR methods. The comparative results between D3Dnet and EDVR are listed in Tables 6 and 7. Note...

评估指标的测试代码（LSNRG,BSF,SCR,CR）以及目标检测代码（tophat,ILCM,IPI的公开源码）已经上传 [Link](https://github.com/XinyiYing/MoCoPnet/tree/main/code/supp)

offset是通过超分辨网络整体的监督loss作为监督信号的，监督D3D能够朝着“达到最佳的超分辨效果”这一目标，不断优化网络参数。 您如果需要继续研究D3D，可以使用一些toy examples做简单的特征可视化。例如，输入图像序列只有一个高亮像素点在缓慢移动，通过观察可视化的特征是否对齐和对齐后可视化特征中高亮像素点的位置，来验证“D3D是否有对齐视频临近帧的作用”这一问题。

Following DCN v1 and v2, we initializes deform_conv_offset by constant zeros. We haven't investigated the effects of different initialization methods for deform_conv_offset learning because constant zeros offset initialization method of...

[SAITD](https://www.scidb.cn/en/detail?dataSetId=808025946870251520&dataSetType=journal)，[Hui](https://www.scidb.cn/en/detail?dataSetId=720626420933459968&dataSetType=journal)， [Anti-UAV](https://anti-uav.github.io/dataset/) 数据集否提供了目标的真值位置。具体参考各个数据集的详细标注。

论文中关于小目标增强的评测都是基于区域邻域计算的，包括SNR gain (SNRG)和contrast gain (CG)，具体公式见论文公式6和公式9。 关于测试集的具体说明在论文的IV-A-1) Datasets， In this paper, we employ the 1st−50th sequences with target annotations of SAITD as the test datasets and the remaining 300 sequences as...

公式6和公式9的分子就是SNR和CR “For simplicity, we only use the best two super-resolved results of D3Dnet and MoCoPnet to perform detection.”这句话的意思是我只用D3Dnet和MoCoPnet在测试集上的超分辨结果进行小目标检测，随后计算检测的结果和原图之间的SNRG BSF SCRG CG四项指标，不是只选择了两张图片。