bconstanzo / phantom

forensic image processing in python

phantom

forensic image processing in python

Requirements

• Python 3.10.x (3.6.x+ still supported, but recommend latest)
• NumPy 1.15+
• scipy 1.2.1+
• matplotlib
• scikit-learn 1.0+
• OpenCV 4.5+
• dlib 19.23+ (19.16 and 19.8 still work)

Installing on Windows

Since Windows tends to be a bit harder to get things working on, we have compiled this instructions to get you going:

1. Get latest Python 3.10.x or 3.11.x (both have been tested and are supported).

2. Install Numpy, Scipy, OpenCV, scikit-learn, and matplotlib (in that order) from Christoph Gohlkes wheels at https://www.lfd.uci.edu/~gohlke/pythonlibs/.

   • Note: these wheels are no longer available. We recomend the same approach as Linux and macOS: pip install -r requirements.txt, but consider that those builds may not be as optimized as Gohlkes wheels were. Things may run a bit slower. If you have available old optimized wheels, prefer those to speed things up (quite) a bit.

3. We've compiled dlib 19.16.0, 19.23.0, and 19.24.2 with AVX for Python 3.6, 3.10, and 3.11 on Windows 64 bits. You can find the corresponding wheels in the /tools directory. Compared to non-AVX version, these are about 2.5 times faster.

   Dlib is developed and maintained by Davis E. King. Check www.dlib.net and https://github.com/davisking/dlib.

4. You can still download the wheel for dlib 19.8.1 from https://pypi.org/simple/dlib/ (thanks to https://github.com/charlielito/install-dlib-python-windows for the pointer). We don't recommend this, and will eventually drop support for older versions of dlib.

Linux and macOS are simpler:

1. Open a console and go to the directory where you've cloned phantom.
2. Run pip install -r requirements.txt.
3. Run pip install ..

You'll need (and most probably already have installed) the platform compiler and it will build a the dependencies as it installs them. This method also works on Windows, but it's less probable that you'll have a proper build environment already set up, and you'll end up with a lower performance installation.

Notes

• When usign our compiled dlib wheel, you may get a puzzling ImportError when trying to import dlib (not phantom). If this happens, you can fix it by installing Intel MKL (download from https://software.intel.com/en-us/mkl) and adding the following values to your environment variables:

  • Add to PATH the following values:

    C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows\redist\intel64_win\tbb\vc_mt;

    C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows\redist\intel64_win\mkl;

    C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows\redist\intel64_win\compiler;

  • LIB (you may have to create it):

    C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows\tbb\lib\intel64_win\vc_mt;

    C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows\mkl\lib\intel64_win;

    C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows\compiler\lib\intel64_win;

  • INCLUDE

    C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows\mkl\include;

    C:\Program Files (x86)\IntelSWTools\compilers_and_libraries\windows\mkl\include\intel64\ilp64;