EDVR is a video restoration framework that uses enhanced deformable convolutions to address challenges in video restoration tasks such as super-resolution and deblurring. The framework includes a Pyramid, Cascading and Deformable (PCD) alignment module and a Temporal and Spatial Attention (TSA) fusion module. The PCD module aligns frames at the feature level using deformable convolutions in a coarse-to-fine manner, while the TSA module applies attention both temporally and spatially to emphasize important features for restoration. EDVR outperforms existing methods in video restoration and enhancement challenges, including the NTIRE19 competition, and demonstrates superior performance on video super-resolution and deblurring tasks. The framework is effective in handling large motions and complex blurring, and it uses a two-stage strategy to further improve performance. EDVR is evaluated on the REDS dataset and shows significant improvements in performance compared to state-of-the-art methods. The framework is also tested on other benchmarks and demonstrates its effectiveness in various video restoration tasks. The results show that EDVR achieves high-quality video restoration with accurate alignment and effective fusion of features.EDVR is a video restoration framework that uses enhanced deformable convolutions to address challenges in video restoration tasks such as super-resolution and deblurring. The framework includes a Pyramid, Cascading and Deformable (PCD) alignment module and a Temporal and Spatial Attention (TSA) fusion module. The PCD module aligns frames at the feature level using deformable convolutions in a coarse-to-fine manner, while the TSA module applies attention both temporally and spatially to emphasize important features for restoration. EDVR outperforms existing methods in video restoration and enhancement challenges, including the NTIRE19 competition, and demonstrates superior performance on video super-resolution and deblurring tasks. The framework is effective in handling large motions and complex blurring, and it uses a two-stage strategy to further improve performance. EDVR is evaluated on the REDS dataset and shows significant improvements in performance compared to state-of-the-art methods. The framework is also tested on other benchmarks and demonstrates its effectiveness in various video restoration tasks. The results show that EDVR achieves high-quality video restoration with accurate alignment and effective fusion of features.