名称错误：K 未定义（即使 K 已定义并使用）

Question

我正在学习 ML/DL，并希望创建一个暹罗神经网络 (SNN) 来识别人脸。

这就是我导入 train.py 文件的方式：

import os
import numpy as np
import cv2
import tensorflow as tf
from keras.models import Model
from keras.layers import Input, Conv2D, MaxPooling2D, Flatten, Dense, Lambda
from keras import backend as K
from keras.optimizers import Adam
from sklearn.model_selection import train_test_split
from utils import create_pairs, load_images_and_labels, save_embeddings

文件 train.py 按预期运行。

但是当我运行我的 recognize.py 文件时。

import cv2
import numpy as np
from keras.models import load_model
from utils import load_embeddings

# Load the trained model and embeddings
siamese_model = load_model('models/siamese_model.h5', compile=False)
registered_embeddings = load_embeddings('models/embeddings.pkl')
base_network = siamese_model.layers[2]  # Extract the base network

# Initialize the face detector
facedetect = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')

# Function to recognize faces in real time
def recognize_face(frame, base_network, registered_embeddings, threshold=0.5):
    faces = facedetect.detectMultiScale(frame, 1.3, 5)
    for x, y, w, h in faces:
        face = frame[y:y+h, x:x+w]
        face = cv2.resize(face, (100, 100))
        face = face.astype('float32') / 255.0
        face = np.expand_dims(face, axis=0)
        
        # Generate embedding for the detected face
        face_embedding = base_network.predict(face)

        # Compare with registered embeddings
        min_dist = float('inf')
        person_name = "Unknown"
        for name, embedding in registered_embeddings.items():
            dist = np.linalg.norm(face_embedding - embedding)
            if dist < min_dist and dist < threshold:
                min_dist = dist
                person_name = name

        # Draw bounding box and name
        cv2.rectangle(frame, (x, y), (x+w, y+h), (0, 255, 0), 2)
        cv2.putText(frame, person_name, (x, y-10), cv2.FONT_HERSHEY_SIMPLEX, 
                    0.9, (255, 255, 255), 2)

    cv2.imshow("Face Recognition", frame)

# Real-time webcam feed
video = cv2.VideoCapture(0)
while True:
    ret, frame = video.read()
    recognize_face(frame, base_network, registered_embeddings)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

video.release()
cv2.destroyAllWindows()

出现此错误： 回溯（最近一次调用最后一次）： 文件“/Users/mac/face_recognition_system/人脸识别系统/recognize.py”，第7行，位于 siamese_model = load_model('models/siamese_model.h5',compile=False) 文件“/Users/mac/face_recognition_system/人脸识别系统/face_recognition/lib/python3.9/site-packages/keras/utils/traceback_utils.py”，第70行，在error_handler中 从 None 引发 e.with_traceback(filtered_tb) 文件“/Users/mac/face_recognition_system/Face Recognition System/train.py”，第 31 行，在 euclidean_distance 中 sumSquared = K.sum(K.square(featsA - featsB), axis=1, keepdims=True)

NameError：调用层“lambda”（Lambda 类型）时遇到异常。 名称“K”未定义

调用“lambda”层接收的参数（Lambda 类型）： • 输入=['tf.Tensor(shape=(None, 4096), dtype=float32)', 'tf.Tensor(shape=(None, 4096), dtype=float32)'] • 掩码=无 • 培训=无

*注意：即使已经安装了tensorflow，我也无法使用tensorflow.keras.X（X是任何东西），因为它说无法解析。 如果有的话，我正在使用 VS 代码。

Answer 1

我通过在 recognize.py 文件中导入 Keras 后端并在文件开头添加 K.clear_session() 来修复此问题。