- 2024-11-12Bracketing Image Restoration and Enhancement with High-Low Frequency Decomposition-高频分解
paper通过密集连接的小卷积核实现细节特征(高频特征提取)提取importtorch.nnasnnclassDense(nn.Module):def__init__(self,in_channels):super(Dense,self).__init__()#self.norm=nn.LayerNorm([in_channels,128,128])#Assuminginputsi
- 2024-10-25DA-CLIP-universal-image-restoration代码详解
DA-CLIP-universal-image-restoration代码详解创建模型model=create_model(opt)device=model.devicecreat_model最终指向ConditionalUNet类,类的主要结构如下:classConditionalUNet(nn.Module):def__init__(self,in_nc,out_nc,nf,ch_mult=[1,
- 2024-05-22Restormer Efficient Transformer for High-Resolution Image Restoration——2022CVPR
大佬链接:Restormer:EfficientTransformerforHigh-ResolutionImageRestoration-知乎(zhihu.com)一.Motivation1.CNN感受野有限,因此无法对长距离像素相关性进行建模;卷积滤波器在推理时具有静态权重,因此不能灵活地适应输入内容2.Transformer模型缓解了CNN的缺点(有限的感
- 2023-12-042023ICCV_FSI Frequency and Spatial Interactive Learning for Image Restoration in Under-Display Camer
三.Network 1. 2.FLB:没看懂是怎么分离的水平和竖直方向 3.SLB:每一层保留一半的通道特征用于细化,其余的在特征重构后输出(没看懂)。Multi-distillationNetwork 超分辨网络的Multi-distillationNetwork(2019ACMMM_LightweightImageSuper-ResolutionwithIn
- 2023-10-31【CVPR2023】Efficient and Explicit Modelling of Image Hierarchies for Image Restoration
>论文:https://readpaper.com/paper/4728855966703960065代码:https://github.com/ofsoundof/GRL-Image-Restoration这个论文的代码地址叫GRL,意思是Global,Regional,Local的意思,作者从三个尺度对特征建模,核心是构建了一个anchoredstripself-attention。如何从Global,R
- 2023-04-20Permutation Restoration (贪心,排序处理) (范围左端点排序,然后取最小点放)
思路:对于每一个bi都会有有一个范围,然后贪心的做,具体的先对这个范围按照左端点排序,然后贪心的去最小的值去放
- 2022-12-07【NeurIPS2022】Cross Aggregation Transformer for Image Restoration
【NeurIPS2022】CrossAggregationTransformerforImageRestoration研究动机:当前方法Transformer方法把图像分成8x8的小块处理,thesquarewindowlacksinter-windo
- 2022-08-14CF EDU 131 D - Permutation Restoration
贪心、扫描线思想D-PermutationRestoration题意有\(1-n\)的一个排列\(a_i\),给定\(b_i\),满足\(b_i=\lfloor\fraci{a_i}\rfloor\),求\(a_i\)(n<=5e5)