Please enable JavaScript.

Coggle requires JavaScript to display documents.

NeRF, Feb 2023: Temporal Interpolation v(x,t) = [vs(x), vd(x,t)] …

- - - - [Jan 2023] HexPlane
        
        Decomposes a 4D spacetime grid into 6 featrue planes (XY, XZ, YZ, XT, YT, ZT)
        
        Quite similar to K-Planes
        
        Feb 2023: Temporal Interpolation
        
        v(x,t) = [vs(x), vd(x,t)]
        
        (v(x,t), d) --> (c, sigma)
        
        [Jan 2023] K-Planes * Planar Factorization
        
        Decomposes a 4D spacetime grid into 6 featrue planes (XY, XZ, YZ, XT, YT, ZT)
  - - - Feb 2023: NeRFPlayer
        
        3 Fields:
        
        Deformation/Stationary Field: (p) --> v_deform, v_static
        
        Newness Field: (p, t) --> v_new
        
        Decomposition Field (p, t) --> (Probabilities)
        
        Last stage (MLP) uses the probabilities with the 3 Vs --> (c, sigma)
  - - - Maybe we can try here different encoding method from the Grid-based branch
  - - - [Nov 2020] Nerfies
        
        2021: D-NeRF
        
        Two networks. 1st (deformation) transfers (x, y, z, t) to canonical frame (x', y', z'). 2nd (canonical) predicts (sigma, c) given (x', y', z')
        
        [2021] DyNeRF
        
        Basically adding t and a 6th input to the MLP
        
        Introduces a keyframe-based training strategy
        
        [May 2021] Dynamic Monocular
        
        [Fab 2022] NeuVV
        Still adds t in with p at the beginig (1 network only)
        
        [Sep 2022] Animated Mesh
        
        1 more item...
        
        [ June 2021] HyperNeRF
  - - - [Sep2022] Time-aware Neural Voxels
        
        [June 2022] Deformable Voxel Grid (NDVG)