#Vimeo90K

UPR-Net employs a unified pyramid recurrent network for efficient video frame interpolation. The model uses lightweight modules and bi-directional flow estimation for robust performance in large motion scenarios with only 1.7M parameters, demonstrating excellence on benchmarks like Vimeo90K. UPR-Net is compatible with PyTorch 1.6 and CUDA 10.2, with options for other versions.