Configuration - juan-cardelino/facial-landmark-detection GitHub Wiki

The configuration is managed by a json file

Path

Direction of all folder path.

Model_dir

Direction of the folder where the face and landmark detection models are stored. Defaul "data"

Input_dir

Direction of the folder where the images to process by [facial-featur-from-file][file] program are stored. Default "input"

Output_dir

Direction of the folder for the outputs of facial-featur-from-video, camera-to-video, roll-angle and yaw-angle programs. Default "output"

Aligned_gir

Direction of the folder for the outputs of the third stage of facial-feature-from-file program. Default "aligned"

Json_dir

Direction of the folder for the outputs of the firts and second stages of facial-feature-from-file program. Default "json"

Pipeline

Independet configuration of all the programs.

From_video

Configuration of facial-feature-from-video program.

Calculate_feature

Condition that controls if facial feature are calculated (calculate_feature = 1) or not (calculate_feature = 0). Default 1

Video_file

Direction and file name of the video to process, if video_file = 0, the video to process is taken from the camera. Default "output/video"

Video_output

Name of the file of the video output. Default "feature-from-video"

Video_detect

Name of the file of the image output. Default "video_detect"

Saving_format

Intager that configurates the output saving format.

1: Video

2: Video and image

3: Image

Default 2.

From_file

Configuration of facial-feature-from-file program.

Etapas

Configurates the number or stages to carry out. Default 3

Minimo_ancho_de_cara

Configurates the minimun width of faces big enough to process. Default 100

Camera_to_video

Configuration of camera-to-video and show-video programs.

Cap_input

Direction where to take the camara cature from. Default 0

Video_output

Name of video output of camera-to-video and video input of show-video. Default "video"

video_length

Intager that restrict the maximum amount of frames in camera-to-video. Default 100

Camera_control

Configuration of roll-angle and yaw-angle programs.

Cap_input

Direction where to take the video cature from. Default 0

Roll_output

Name of video output of roll-angle. Default "roll_angle"

Yaw_output

Name of video output of yaw-angle. Default "yaw_angle"

Binarizar_dataset

Configuration of reducir-y-binarizar-dataset, control-database-binarizada and control-database-continuous programs.

Dataset_dir

Direction of the input dataset jsons to reducir-y-binarizar-dataset, control-database-headpose and control-database-continuous. Default "FFHQ_json_features"

Dataset_binarizada

Direction and name of the output of reducir-y-binarizar-dataset and input of control-database-binarizada. Default "ffhq-feature-dataset-binarizado"

Color_dataset

Configuration of reducir-y-color-dataset and control-database-color programs.

Dataset_input

Direction and name of the input dataset to reducir-y-color-dataset. Default "ffhq-dataset-v2"

Dataset_output

Direction and name of the output of reducir-y-color-dataset and input of control-database-color. Default "ffhq-dataset-color"

General

General configuration for many programs and modules

Resize

Configurates the size of the images and videos shown in screen. Default [1920, 1080] (full screen)

Face detection model

Name of the face detection model file. Deafult "haarcascade_frontalface_alt2.xml"

Face detection model_url

URL link to download the face detection model. Default "https://raw.githubusercontent.com/opencv/opencv/master/data/haarcascades/haarcascade_frontalface_alt2.xml"

Landmark detection model

Name of the landmark detection model file. Default "LFBmodel.yaml"

Landmark detection model_url

URL link to download the face detection model. Default "https://github.com/kurnianggoro/GSOC2017/raw/master/data/lbfmodel.yaml"

Json_suffix_detect

Suffix of the json files that have face and landmark detection. Default "detection'

Json_suffix_data

Suffix of the json files that have facial features calculations. Default "data"