Scene-Adaptive Video Frame Interpolation via Meta-Learning

@article{Choi2020SceneAdaptiveVF,
  title={Scene-Adaptive Video Frame Interpolation via Meta-Learning},
  author={Myungsub Choi and Janghoon Choi and Sungyong Baik and Tae Hyun Kim and Kyoung Mu Lee},
  journal={2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2020},
  pages={9441-9450}
}
Video frame interpolation is a challenging problem because there are different scenarios for each video depending on the variety of foreground and background motion, frame rate, and occlusion. It is therefore difficult for a single network with fixed parameters to generalize across different videos. Ideally, one could have a different network for each scenario, but this is computationally infeasible for practical applications. In this work, we propose to adapt the model to each video by making… 

Figures and Tables from this paper

Test-Time Adaptation for Video Frame Interpolation via Meta-Learning

This work proposes MetaVFI, an adaptive video frame interpolation algorithm that uses additional information readily available at test time but has not been exploited in previous works to obtain significant performance gains with only a single gradient update without introducing any additional parameters.

Motion-Aware Dynamic Architecture for Efficient Frame Interpolation

This work proposes an effective framework that assigns varying amounts of computation for different regions of a video frame interpolation system, and demonstrates that the proposed framework can significantly reduce the computation cost (FLOPs) while maintaining the performance.

Multiple Video Frame Interpolation via Enhanced Deformable Separable Convolution

A novel non-flow kernel-based approach that is referred to as enhanced deformable separable convolution (EDSC) to estimate not only adaptive kernels, but also offsets, masks and biases to make the network obtain information from non-local neighborhood.

Revisiting Adaptive Convolutions for Video Frame Interpolation

This work shows, somewhat surprisingly, that it is possible to achieve near state-of-the-art results with an older, simpler approach, namely adaptive separable convolutions, by a subtle set of low level improvements.

Training Weakly Supervised Video Frame Interpolation with Events

This work shows event-based frame interpolation can be trained without the need of high frame-rate videos via a novel weakly supervised framework that corrects image appearance by extracting complementary information from events and supplants motion dynamics modeling with attention mechanisms.

Splatting-based Synthesis for Video Frame Interpolation

A deep learning approach that solely relies on splatting to synthesize interpolated frames for video frame interpolation is proposed, which is not only much faster than similar approaches, especially for multi-frame interpolation, but can also yield new state-of-the-art results at high resolutions.

Beyond a Video Frame Interpolator: A Space Decoupled Learning Approach to Continuous Image Transition

This work rethink the VFI problem and formulate it as a continuous image transition (CIT) task, whose key issue is to transition an image from one space to another space continuously, and proposes space decoupled learning (SDL), a general-purpose solution to the CIT problem.

Context-based video frame interpolation via depthwise over-parameterized convolution

Experimental results demonstrate that the proposed context-based video frame interpolation method via depthwise over-parameterized convolution performs qualitatively and quantitatively better than state-of-the-art methods.

Video Restoration Framework and Its Meta-adaptations to Data-Poor Conditions

This work proposes a generic architecture that is effective for any weather condition due to the ability to extract robust feature maps without any domain-specific knowledge and shows comprehensive results on video de-hazing and de-raining datasets in addition to the meta-learning based adaptation results on night-time video restoration tasks.

Long-term Video Frame Interpolation via Feature Propagation

A propagation network (PNet) is proposed by extending the classic feature-level forecasting with a novel motion-to-feature approach that can safely propagate from one side of the input up to a reliable time frame using the other input as a reference when there is a large gap between inputs.

References

SHOWING 1-10 OF 49 REFERENCES

PhaseNet for Video Frame Interpolation

This work proposes a new approach, PhaseNet, that is designed to robustly handle challenging scenarios while also coping with larger motion, and shows that this is superior to the hand-crafted heuristics previously used in phase-based methods and compares favorably to recent deep learning based approaches for video frame interpolation on challenging datasets.

Channel Attention Is All You Need for Video Frame Interpolation

A simple but effective deep neural network for video frame interpolation, which is end-to-end trainable and is free from a motion estimation network component, and achieves outstanding performance compared to the existing models with a component for optical flow computation.

Video Frame Synthesis Using Deep Voxel Flow

This work addresses the problem of synthesizing new video frames in an existing video, either in-between existing frames (interpolation), or subsequent to them (extrapolation), by training a deep network that learns to synthesize video frames by flowing pixel values from existing ones, which is called deep voxel flow.

Super SloMo: High Quality Estimation of Multiple Intermediate Frames for Video Interpolation

This work proposes an end-to-end convolutional neural network for variable-length multi-frame video interpolation, where the motion interpretation and occlusion reasoning are jointly modeled.

Video Frame Interpolation via Adaptive Convolution

This paper presents a robust video frame interpolation method that considers pixel synthesis for the interpolated frame as local convolution over two input frames and employs a deep fully convolutional neural network to estimate a spatially-adaptive convolution kernel for each pixel.

Unsupervised Video Interpolation Using Cycle Consistency

This work proposes unsupervised techniques to synthesize high frame rate videos directly from low frame rate films using cycle consistency, and introduces a pseudo supervised loss term that enforces the interpolated frames to be consistent with predictions of a pre-trained interpolation model.

Deep Video Frame Interpolation Using Cyclic Frame Generation

A new loss term, the cycle consistency loss, which can better utilize the training data to not only enhance the interpolation results, but also maintain the performance better with less training data is introduced.

Quadratic video interpolation

This work proposes a quadratic video interpolation method which exploits the acceleration information in videos, allows prediction with curvilinear trajectory and variable velocity, and generates more accurate interpolation results.

Video Frame Interpolation via Adaptive Separable Convolution

This paper develops a deep fully convolutional neural network that takes two input frames and estimates pairs of 1D kernels for all pixels simultaneously, which allows for the incorporation of perceptual loss to train the neural network to produce visually pleasing frames.

Context-Aware Synthesis for Video Frame Interpolation

  • Simon NiklausFeng Liu
  • Computer Science
    2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition
  • 2018
A context-aware synthesis approach that warps not only the input frames but also their pixel-wise contextual information and uses them to interpolate a high-quality intermediate frame and outperforms representative state-of-the-art approaches.