README.md

pico-detect

This library is a reimplementation of Pixel Intensity Comparison-based Object (PICO) detection algorithms in Rust:

• Detector: Cascade of binary classifiers from pico;
• Localizer: Localization with an ensemble of randomized trees from picojs (see lploc.js);
• Shaper: Alignment with an ensemble of regression trees from dlib (see shape_predictor).

Example

To run CLI example, which takes an image, finds all faces, detects some landmarks and pupils:

NOTE: Git LFS is needed to resolve binary files with git clone.

If you don't want to use Git LFS you can download models (and test image) direct from this repo (see model column in the table below) and put them under models/ directory.

cargo run --release --example detect-faces -- --models-dir models -i "assets/test.png" --score 35.0 -o result.png

Output image result.png should be like this:

Models

Each algorithm requires to be loaded with correspondent binary model.

model	algorithm	source	Description
face.detector.bin	Detector	pico	Human face classifier
pupil.localizer.bin	Localizer	puploc	Human eye pupil localizer
face-5.shaper.bin	Shaper	shape_predictor_5_face_landmarks	Human 5 face landmarks

References

1. N. Markus, M. Frljak, I. S. Pandzic, J. Ahlberg and R. Forchheimer, "Object Detection with Pixel Intensity Comparisons Organized in Decision Trees"

2. Eye pupil localization with an ensemble of randomized trees

3. One Millisecond Face Alignment with an Ensemble of Regression Trees