Official PyTorch implementation of Scalable Neural Video Representations with Learnable Positional Features (NeurIPS 2022). - subin-kim-cv/NVP
we tested for encoding position, direction and combinations thereof (Fig.2c). We find that position plus direction is the most informative label29(Fig.2cand Extended Data Fig.5a–d). This is evident both in the embedding and the value of the loss function on convergence, which serves ...