Faceless

Faceless analyzes images and video with Ultralytics YOLO, then moves media that does not contain both a face (class 264) and at least one required match class (default: 216,594). Non-matching files land in the output directory (default: faceless) and are grouped into label-named subfolders when labels are available.

Install

Download the Windows exe or wheel from the GitHub releases page: https://github.com/mr-szgz/faceless/releases.

Install a wheel directly:

pip install ./faceless-0.7.1-py3-none-any.whl --force

Or clone the repo and install in editable mode:

pip install -e .

Usage

Move non-matching media (grouped by label when available):

faceless "p:/path/to/media"

Show help:

$ faceless --help                                                                                                            
usage: faceless [-h] [-Force] [-Confidence CONF_FLOAT] [-RequireIds YOLO_CLASS_INTS] [-Directory OUTPUT_DIR] [path]                                              

positional arguments:
  path                  Source directory containing videos/images

options:
  -h, --help            show this help message and exit
  -Force, --force       Force regeneration of labels
  -Confidence CONF_FLOAT, --confidence CONF_FLOAT
                        Model confidence threshold
  -RequireIds YOLO_CLASS_INTS, --require-ids YOLO_CLASS_INTS
                        YOLO class IDs to keep comma-separated. All classes available in faceless/datasets/OpenImagesV7.yaml. Default: "216,594"
  -Directory OUTPUT_DIR, --directory OUTPUT_DIR
                        Override output directory. Default: ./faceless

Windows Context Menu

If you downloaded run-faceless.exe from a release, you can install or remove the Explorer context menu entry:

run-faceless --install
run-faceless --uninstall
run-faceless --info

Changes

See CHANGELOG.md for release notes and version history.