Deep Bayesian Video Frame Interpolation

Abstract. We present deep Bayesian video frame interpolation, a novel approach for upsampling a low frame-rate video temporally to its higher frame-rate counterpart. Our approach learns posterior distributions of optical flows and frames to be interpolated, which is optimized via learned gradient descent for fast convergence. Each learned step is a lightweight network manipulating gradients of the log-likelihood of estimated frames and flows. Such gradients, parameterized either explicitly or implicitly, model the fidelity of current estimations when matching real image and flow distributions to explain the input observations. With this approach we show new records on 8 of 10 benchmarks, using an architecture with half the parameters of the state-of-the-art model.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Video Frame Interpolation DAVIS DBVI PSNR 28.61 # 1
SSIM 0.905 # 1
Video Frame Interpolation GoPro DBVI PSNR 31.73 # 1
SSIM 0.947 # 1
Video Frame Interpolation SNU-FILM (easy) DBVI PSNR 40.46 # 2
SSIM 0.991 # 2
Video Frame Interpolation SNU-FILM (extreme) DBVI PSNR 25.90 # 1
SSIM 0.876 # 1
Video Frame Interpolation SNU-FILM (hard) DBVI PSNR 31.68 # 2
SSIM 0.953 # 1
Video Frame Interpolation SNU-FILM (medium) DBVI PSNR 36.95 # 2
SSIM 0.985 # 1
Video Frame Interpolation X4K1000FPS DBVI PSNR 32.89 # 1
SSIM 0.939 # 1

Methods


No methods listed for this paper. Add relevant methods here