face3d: Python tools for processing 3D face

Introduction

Hope this helps & you enjoy it. ^_^

Structure

# Since triangle mesh is the most popular representation of 3D face, 
# the main part is mesh processing.
mesh/             # written in python(numpy)
|  io.py                 # read & write obj
|  vis.py                # plot mesh
|  transform.py          # transform mesh & estimate matrix
|  light.py              # add light & estimate light(to do)
|  render.py             # obj to image using rasterization render

# When can not use vercorization to optimize, loops in python are too slow to use. 
# So write core parts(loops) in c++, then use cython to compile them for python api.
mesh_cython/      # written in c++(cython)
|  render.py             # the same API as in mesh/render.py, with faster speed.
|  light.py              # the same API as in mesh/light.py, with faster speed.
|  mesh_core*            # c++ codes of core parts.

# 3DMM is one of the most popular methods to generate & reconstruct 3d face.
morphabel_model/  # witten in python(numpy)
|  morphable_model.py    # morphable model class: generate & fit
|  fit.py                # estimate shape&expression parameters. 3dmm fitting.
|  load.py               # load 3dmm data

Examples:

cd ./examples

3dmm. python 2_3dmm.py

left: random example generated by 3dmm

right: fitting face with 3dmm using 68 key points
transform. python 3_transform.py
left:

fix camera position & use orthographic projection. (often used in reconstruction)

then transform face object: scale, change pitch angle, change yaw angle, change roll angle

right:

fix obj position & use perspective projection(fovy=30). (simulating real views)

then move camera position and rotate camera: from far to near, down & up, left & right, rotate camera
light. python 4_light.py

single point light: from left to right, from up to down, from near to far
image map python 6_image_map.py

render different attributes in image pixels.

: depth, pncc, uv coordinates
uv map python 7_uv_map.py

render different attributes in uv coordinates.

: colors(texture map), position(2d facial image & corresponding position map)

Getting Started

Prerequisite

Python 2 or Python 3
Python packages:
- numpy
- skimage (for reading&writing image)
- scipy (for loading mat)
- matplotlib (for show)
- Cython (for compiling c++ file)

Usage

Clone the repository

git clone https://github.com/YadiraF/face3d
cd face3d

Compile c++ files to .so for python use (ignore if you use numpy version)
```
cd face3d/mesh_cython
python setup.py build_ext -i 
```
Prepare BFM Data (ignore if you don't use 3dmm)

see Data/BFM/readme.md
Run examples

(examples use cython version, you can change mesh_cython into mesh to use numpy version)
```
cd examples
python 1_pipeline.py 
```

zhoupan9109/face3d

face3d: Python tools for processing 3D face

Introduction

Structure

Examples:

Getting Started

Prerequisite

Usage