Sign-Language/test_classifier.py at main · jaovitm/Sign-Language · GitHub

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
import cv2
import mediapipe as mp
import os
import pickle
import numpy as np
import warnings

# Suprimir warnings específicos do protobuf
warnings.filterwarnings("ignore", category=UserWarning, module="google.protobuf.symbol_database")

cap = cv2.VideoCapture(0)

mp_hands = mp.solutions.hands
mp_drawing = mp.solutions.drawing_utils
mp_drawing_styles = mp.solutions.drawing_styles

# Set max_num_hands=1 for single hand detection
hands = mp_hands.Hands(static_image_mode=False, min_detection_confidence=0.3)

model_dict = pickle.load(open('model.pickle', 'rb'))
model = model_dict['model']

# Check how many features the model expects
print(f"Model expects {model.n_features_in_} features")

labels_dict = {0: 'C', 1: 'M', 2: 'R', 3: 'V'}

while True:
    data_aux = []
    x_ = []
    y_ = []

    ret, img = cap.read()

    if not ret:
        break

    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    result = hands.process(img_rgb)

    if result.multi_hand_landmarks:
        for hand_landmarks in result.multi_hand_landmarks:
            mp_drawing.draw_landmarks(
                img,
                hand_landmarks,
                mp_hands.HAND_CONNECTIONS,
                mp_drawing_styles.get_default_hand_landmarks_style(),
                mp_drawing_styles.get_default_hand_connections_style())

            # Extract features from single hand
            for landmark in hand_landmarks.landmark:
                x_.append(landmark.x)
                y_.append(landmark.y)
                data_aux.extend([landmark.x, landmark.y])

        # Pad features to match training data (84 features expected)
        if len(data_aux) > 0:
            try:
                # Pad the features to 84 (same as training)
                features = np.array(data_aux)
                if len(features) < 84:
                    # Pad with zeros to reach 84 features
                    padding = np.zeros(84 - len(features))
                    features = np.concatenate([features, padding])
                elif len(features) > 84:
                    # Truncate to 84 features if somehow we have more
                    features = features[:84]

                prediction = model.predict([features])
                predicted_character = labels_dict[int(prediction[0])]

                # Visual feedback
                if x_ and y_:
                    H, W, _ = img.shape
                    x1 = int(min(x_) * W) - 10
                    y1 = int(min(y_) * H) - 10
                    x2 = int(max(x_) * W) - 10
                    y2 = int(max(y_) * H) - 10

                    cv2.rectangle(img, (x1, y1), (x2, y2), (0, 0, 0), 4)
                    cv2.putText(img, predicted_character, (x1, y1 - 10),
                              cv2.FONT_HERSHEY_SIMPLEX, 1.3, (0, 0, 0), 3, cv2.LINE_AA)

                print(f"Predicted: {predicted_character}")

            except Exception as e:
                print(f"Prediction error: {e}")

    cv2.imshow('Sign Language Detection', img)

    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()