Skip to content

Latest commit

 

History

History
64 lines (35 loc) · 1.95 KB

README.md

File metadata and controls

64 lines (35 loc) · 1.95 KB
Raw

Computational caricaturization of surfaces

This is an unofficial reproduction of "Computational caricaturization of surfaces" by Martan Sela, Yonathan Aflalo, and Ron Kimmmel. (CVIU 2015).

Building

Requirement

  • Eigen3
  • CMake

Build Command

git clone <URL> --recursive <dir>

cd <dir> && mkdir build && cmake -DCMAKE_BUILD_TYPE=Release ../ && make

Demo executable

Usage 1: <executable> ref <reference .obj file> <target .obj file> <beta (default 0.2)> <output .obj path (default: output.obj)>
Usage 2: <executable> noref <target .obj file> <gamma (default 0.2)> <output .obj path (default: output.obj)>

For the semantics of the exaggeration factors beta and gamma, please refer to the paper. 0.0 means no exaggeration at all. Put values somewhere between 0.0 to 0.8.

Demo

Without reference

Input

image

Result with gamma = 0.2

image

Result with gamma = 0.3

image

With reference

Reference

image

Input

image

Result with beta = 0.6

image

Discussion

  • Although the algorithm may give us interesting shapes, this algorithm largely modifies global scale of input mesh (i.e. the size of the bounding box). There may be some possible improvements.
  • The formulation for with-reference case is modified. (at src/compcari.cpp:131, negative sign before std::log is added). This seems like a typo of the paper.

Disclaimer

The meshes used in the examples are generated using the Basel face model (Paysan, P., Knothe, R., Amberg, B., Romdhani, S., & Vetter, T. (2009, September). A 3D face model for pose and illumination invariant face recognition. In 2009 Sixth IEEE International Conference on Advanced Video and Signal Based Surveillance (pp. 296-301). Ieee.). The LICENSE only covers my code (src/ include/ demo/), not example meshes.