app.py

#!/usr/bin/env python

from __future__ import annotations

import os
import pathlib
import sys

import cv2
import gradio as gr
import numpy as np
import torch

sys.path.insert(0, "face_detection")
sys.path.insert(0, "face_alignment")

from ibug.face_alignment import FANPredictor
from ibug.face_detection import RetinaFacePredictor

DESCRIPTION = "# [ibug-group/face_alignment](https://github.com/ibug-group/face_alignment)"

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
detector = RetinaFacePredictor(threshold=0.8, device=device, model=RetinaFacePredictor.get_model("mobilenet0.25"))
model_names = [
    "2dfan2",
    "2dfan4",
    "2dfan2_alt",
]
models = {name: FANPredictor(device=device, model=FANPredictor.get_model(name)) for name in model_names}


def predict(image: np.ndarray, model_name: str, max_num_faces: int, landmark_score_threshold: int) -> np.ndarray:
    model = models[model_name]

    # RGB -> BGR
    image = image[:, :, ::-1]

    faces = detector(image, rgb=False)
    if len(faces) == 0:
        raise RuntimeError("No face was found.")
    faces = sorted(list(faces), key=lambda x: -x[4])[:max_num_faces]
    faces = np.asarray(faces)
    landmarks, landmark_scores = model(image, faces, rgb=False)

    res = image.copy()
    for face, pts, scores in zip(faces, landmarks, landmark_scores):
        box = np.round(face[:4]).astype(int)
        cv2.rectangle(res, tuple(box[:2]), tuple(box[2:]), (0, 255, 0), 2)
        for pt, score in zip(np.round(pts).astype(int), scores):
            if score < landmark_score_threshold:
                continue
            cv2.circle(res, tuple(pt), 2, (0, 255, 0), cv2.FILLED)

    return res[:, :, ::-1]


examples = [[path.as_posix(), model_names[0], 10, 0.2] for path in pathlib.Path("images").rglob("*.jpg")]

with gr.Blocks(css="style.css") as demo:
    gr.Markdown(DESCRIPTION)
    with gr.Row():
        with gr.Column():
            image = gr.Image(type="numpy", label="Input")
            model_name = gr.Radio(model_names, type="value", value=model_names[0], label="Model")
            max_num_faces = gr.Slider(1, 20, step=1, value=10, label="Max Number of Faces")
            landmark_score_thrshold = gr.Slider(0, 1, step=0.05, value=0.2, label="Landmark Score Threshold")
            run_button = gr.Button()
        with gr.Column():
            result = gr.Image(label="Output")
    gr.Examples(
        examples=examples,
        inputs=[image, model_name, max_num_faces, landmark_score_thrshold],
        outputs=result,
        fn=predict,
        cache_examples=os.getenv("CACHE_EXAMPLES") == "1",
    )
    run_button.click(
        fn=predict,
        inputs=[image, model_name, max_num_faces, landmark_score_thrshold],
        outputs=result,
        api_name="predict",
    )


if __name__ == "__main__":
    demo.queue(max_size=20).launch()