Commit 0398b2aa authored by Richard Torenvliet's avatar Richard Torenvliet

Update readme

parent c09fab29
# About this tool
This tool is meant to as a pipeline to show 2D and 3D reconstructions using An
Active Appearance Model and a 2D or 3D model to rebuild the face. Face
reconstruction is a difficult subject but like with everything, if you
understand the steps, it's actually ok. This small library, can give you a
feeling what is needed to solve this problem, but also some quick-and-dirty
tricks are used. Like using dlib to solve landmark detection, instead of
estimating them using a more traditional way by as done by [[Coots|coots]].
Instead, dlib uses a sophistaticated approach to estimate 2D landmarks. This
capability is re-used to find the PCA parameters needed to rebuild a person's
face.
## Prerequisites ## Prerequisites
Run the following command - Docker
~~~~
$ make
$ source bin/activate
$ make show_reconstruction
~~~~
Will get the data needed for this tool. # Run
- $ make
> *Note*: this will build the docker image, retrieves the imm dataset and dlib
> trained landmark file.
> $ make server
- Use https://github.com/icyrizard/py-3d-face-reconstruction-viewer to run a
viewer which uses the server to make the reconstruction insightful.
# Imm dataset # Imm dataset
# IBUG + dlib # IBUG + dlib
For the IBUG dataset we use dlib to detect landmarks. You will need the train For the IBUG dataset we use dlib to detect landmarks. You will need the train
file for that. http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2 file for that. http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2
...@@ -22,3 +34,4 @@ $ bzip2 -d shape_predictor_68_face_landmarks.dat.bz2 ...@@ -22,3 +34,4 @@ $ bzip2 -d shape_predictor_68_face_landmarks.dat.bz2
## References ## References
1. [imm_dataset](http://www.imm.dtu.dk/~aam/datasets/datasets.html, "Imm dataset") 1. [imm_dataset](http://www.imm.dtu.dk/~aam/datasets/datasets.html, "Imm dataset")
2. [coots](https://www.cs.cmu.edu/~efros/courses/AP06/Papers/cootes-eccv-98.pdf, "Coots")
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment