Human Motion and Novel View Synthesis Compilation

What is this?

This project page is a collection of papers (and possibly code) that has something to do with human motion and novel view synthesis in 2D and 3D space. There are two major classes of methods when doing it in 2D space:

Spatial Transformation (using flow information, etc)
Conditioned Generation (pose label map, etc)
Texture-space and Geometry generation (UV correspondence, texture generation, using 3D, etc)
Feature-space manipulation (Spatial attention, kernels, latent embeddings, etc)

Papers are also split into several sub-categories (ones with the project page):

Novel View Synthesis - This category is about rendering/generating an image of a human within another pose, may be a single picture or a video
Motion Retargeting, Representation, Continuation - This category is about the modeling of human motion as well as skeletal prediction. May also include temporally-coherent pose estimation
3D Human Digitization - From a(n ideally) single image, this category deals with 3D human generation along with texture and/or shape inference from unseen angles

Papers that do not have first-party implementations are put into another category with two sub-categories:

Architectural - Contribution is mainly about architecture
Representation - Contribution is related to information representation, spatial representation, or feature extraction methods

This project page is only intended to supplement my studies and research where everything is (hopefully nicely) organized into one place.

Papers with Project Pages