prep

Preparing Face Crops

Morphace handles size mismatches, but it cannot make every composition line up. A close-up portrait and a full-body distance shot can both contain detectable faces, but the morph will spend too much motion on background and body area. (Refer to the gallery for examples)

Run those images through morphace-prep when you want the transition to focus on faces. The prep command detects every face in each source image, calculates an alignment quadrilateral from the 68 landmarks, prepares the source canvas, and warps each face into a square PNG.

Prep Image Example

The example below shows an image before and after being processed.

First source portrait used in the Morphace demo — Source image

Prepared square face crop from the Morphace demo source image — Prepared image

Usage

usage: morphace prep [-f] [-s OUTPUT_SIZE] [-x X_SCALE] [-y Y_SCALE]
                     [-e EM_SCALE] [-a] [-l LANDMARK_MODEL] [--debug] [-h]
                     [-V]
                     source [aligned_dir]

Detect faces in raw images and prepare aligned square PNG crops for morphing.

positional arguments:
  source                Path to a directory of images or a single image file
  aligned_dir           Directory for aligned images (default:
                        <source>/cropped for directories, same directory as
                        source for files)

output options:
  -f, --force           Overwrite existing aligned images (default: False)

alignment options:
  -s, --output-size OUTPUT_SIZE
                        Pixel dimension of the output square crop (default:
                        1024)
  -x, --x-scale X_SCALE
                        Horizontal crop extent around face; >1.0 includes more
                        context, <1.0 crops tighter (default: 1.0)
  -y, --y-scale Y_SCALE
                        Vertical crop extent around face; >1.0 includes more
                        context, <1.0 crops tighter (default: 1.0)
  -e, --em-scale EM_SCALE
                        Shift crop center from eyes toward mouth (default:
                        0.1)
  -a, --alpha           Use alpha channel for padded regions instead of
                        reflected padding (default: False)

model options:
  -l, --landmark-model LANDMARK_MODEL
                        Path to shape_predictor_68_face_landmarks_GTX.dat
                        model file. If omitted, morphace checks
                        MORPHACE_LANDMARK_MODEL and then the default user data
                        directory. (default: None)

diagnostic options:
  --debug               Enable debug-level logging (default: False)

general options:
  -h, --help            show this help message and exit
  -V, --version         show program's version number and exit

Example Usage

Prepare a directory of raw images:

morphace prep /path/to/images/ /path/to/save/images

Prepare a single image and save the prepared image to the same directory:

morphace prep /path/to/image.jpg