Multi-Scale Video Frame-Synthesis Network with Transitive Consistency Loss

Traditional approaches to interpolating/extrapolating frames in a video sequence require accurate pixel correspondences between images, e.g., using optical flow. Their results stem on the accuracy of optical flow estimation, and would generate heavy artifacts when flow estimation failed. Recent methods using auto-encoder have shown impressive progress, however they are usually trained for specific interpolation/extrapolation settings and lack of flexibility and generality for more applications. Moreover, these models are usually heavy in terms of model size which constrains applications on mobile devices. In order to reduce these limitations, we propose a unified network to parameterize the interest frame position and therefore infer interpolated/extrapolated frames within the same framework. To better regularize the network, we introduce a transitive consistency loss and train the network with adversarial training. We adopt a multi-scale structure for the network so that the parameters can be shared across multi-layers. Our approach avoids expensive global optimization of optical flow methods, and is efficient and flexible for video interpolation/extrapolation applications. Experimental results have shown that our method performs favorably against state-of-the-art methods.