File size: 4,437 Bytes
fc1a577
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
import cv2
import numpy as np
from onnxruntime import InferenceSession

import box_utils_numpy
from auto_rotate import align_face


def softmax(x):
    e_x = np.exp(x - np.max(x))
    return e_x / e_x.sum(axis=0)


def crop_square(img, size, interpolation=cv2.INTER_AREA):
    h, w = img.shape[:2]
    min_size = np.amin([h, w])

    # Centralize and crop
    crop_img = img[
        int(h / 2 - min_size / 2) : int(h / 2 + min_size / 2),
        int(w / 2 - min_size / 2) : int(w / 2 + min_size / 2),
    ]
    resized = cv2.resize(crop_img, (size, size), interpolation=interpolation)

    return resized


class SpoofyNet:
    def __init__(self):
        self.face_model = InferenceSession("models/slim-facedetect.onnx")
        self.face_inputname = self.face_model.get_inputs()[0].name
        self.classifier = InferenceSession("models/spoof.onnx")

    def find_boxes(
        self,
        width,
        height,
        confidences,
        boxes,
        prob_threshold,
        iou_threshold=0.3,
        top_k=-1,
    ):
        boxes = boxes[0]
        confidences = confidences[0]
        picked_box_probs = []
        picked_labels = []
        for class_index in range(1, confidences.shape[1]):
            probs = confidences[:, class_index]
            mask = probs > prob_threshold
            probs = probs[mask]
            if probs.shape[0] == 0:
                continue
            subset_boxes = boxes[mask, :]
            box_probs = np.concatenate([subset_boxes, probs.reshape(-1, 1)], axis=1)
            box_probs = box_utils_numpy.hard_nms(
                box_probs,
                iou_threshold=iou_threshold,
                top_k=top_k,
            )
            picked_box_probs.append(box_probs)
            picked_labels.extend([class_index] * box_probs.shape[0])
        if not picked_box_probs:
            return np.array([]), np.array([]), np.array([])
        picked_box_probs = np.concatenate(picked_box_probs)
        picked_box_probs[:, 0] *= width
        picked_box_probs[:, 1] *= height
        picked_box_probs[:, 2] *= width
        picked_box_probs[:, 3] *= height
        return (
            picked_box_probs[:, :4].astype(np.int32),
            np.array(picked_labels),
            picked_box_probs[:, 4],
        )

    def tta(self, src):
        horizontal_rot = cv2.rotate(src, cv2.ROTATE_180)
        grayscale = cv2.cvtColor(src, cv2.COLOR_RGB2GRAY)
        grayscale = cv2.cvtColor(grayscale, cv2.COLOR_GRAY2RGB)
        return [src, horizontal_rot, grayscale]

    def find_spoof(self, img):
        ret = []
        threshold = 0.6
        image_mean = np.array([127, 127, 127])

        image = cv2.resize(img, (320, 240))
        image = (image - image_mean) / 128
        image = np.transpose(image, [2, 0, 1])
        image = np.expand_dims(image, axis=0)
        image = image.astype(np.float32)

        confidences, boxes = self.face_model.run(None, {self.face_inputname: image})
        boxes, _, _ = self.find_boxes(
            img.shape[1], img.shape[0], confidences, boxes, threshold
        )

        classify_mean, classify_std = [0.485, 0.456, 0.406], [0.229, 0.224, 0.225]
        for i in range(boxes.shape[0]):
            (startX, startY, endX, endY) = boxes[i, :]

            face = img[startY:endY, startX:endX]
            if face.size == 0:
                continue
            face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)

            # Preprocess
            face = align_face(face)
            face = crop_square(face, 256)

            probs_all = []
            for face in self.tta(face):
                # Normalize
                face = face / 255.0
                face = (face - classify_mean) / classify_std
                face = np.transpose(face, [2, 0, 1])
                face = np.expand_dims(face, axis=0)
                face = face.astype(np.float32)

                predicted = self.classifier.run(None, {"input": face})
                predicted_id = np.argmax(predicted)
                probs = softmax(predicted[0][0])
                probs_all.append(probs)

            final_probs = np.mean(probs_all, axis=0)
            predicted_id = np.argmax(final_probs)
            ret.append(
                {
                    "coords": (startX, startY, endX, endY),
                    "is_real": bool(predicted_id),
                    "probs": final_probs[predicted_id],
                }
            )
        return ret