Issue reading the image using cv2.imread() when training my model

Question

import os
import cv2
import numpy as np
from sklearn.model_selection import StratifiedKFold
from keras.utils import to_categorical
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout, BatchNormalization
import tensorflow.keras as keras

# Define the path to the FER-2013 dataset directory
data_path = "/home/bwc/an561/Dissertation-Project/dataset"

# Define the list of emotions and their corresponding labels
emotion_labels = {
    'angry': 0,
    'disgust': 1,
    'fear': 2,
    'happy': 3,
    'neutral': 4,
    'sad': 5,
    'surprise': 6
}

# Load the images and labels from the train dataset
train_images = []
train_labels = []

train_folder_path = os.path.join(data_path, "/train")

for emotion_folder in os.listdir(train_folder_path):
    emotion = emotion_folder
    emotion_folder_path = os.path.join(train_folder_path, emotion_folder)

    # Skip non-directory files
    if not os.path.isdir(emotion_folder_path):
        continue

    for image_filename in os.listdir(emotion_folder_path):
        # Skip non-image files
        if not image_filename.endswith(('.jpg', '.jpeg', '.png')):
            continue

        image_path = os.path.join(emotion_folder_path, image_filename)
        image = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE)
        image = cv2.resize(image, (48, 48))  # Resize images to the desired shape
        train_images.append(image)
        train_labels.append(emotion_labels[emotion])
# Convert train images and labels to numpy arrays
train_images = np.array(train_images)
train_labels = np.array(train_labels)

# Normalize pixel values to range between 0 and 1
train_images = train_images / 255.0

# Perform one-hot encoding on the train labels
train_labels = to_categorical(train_labels)

# Define the model
input_shape = (48, 48, 1)  # Assuming input images are grayscale and have size 48x48
model = Sequential()
model.add(Conv2D(64, (3, 3), activation='relu', input_shape=input_shape))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(BatchNormalization())
model.add(Dropout(0.25))
model.add(Conv2D(128, (3, 3), activation='relu'))
model.add(Conv2D(128, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(BatchNormalization())
model.add(Dropout(0.25))
model.add(Conv2D(256, (3, 3), activation='relu'))
model.add(Conv2D(256, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(BatchNormalization())
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(512, activation='relu'))
model.add(BatchNormalization())
model.add(Dropout(0.5))
model.add(Dense(256, activation='relu'))
model.add(BatchNormalization())
model.add(Dropout(0.5))
model.add(Dense(7, activation='softmax'))

# Compile the model with a lower learning rate
learning_rate = 0.001
optimizer = keras.optimizers.Adam(learning_rate=learning_rate)
model.compile(optimizer=optimizer, loss='categorical_crossentropy', metrics=['accuracy'])

# Perform cross-validation
k_folds = 5
skf = StratifiedKFold(n_splits=k_folds, shuffle=True, random_state=42)

accuracies = [] 
for fold, (train_index, val_index) in enumerate(skf.split(train_images, np.argmax(train_labels, axis=1))):
    print(f"Training fold {fold + 1}/{k_folds}")

    # Split the data into training and validation sets
    X_train, X_val = train_images[train_index], train_images[val_index]
    y_train, y_val = train_labels[train_index], train_labels[val_index]

    # Train the model
    epochs = 20
    model.fit(X_train, y_train, validation_data=(X_val, y_val), epochs=epochs, batch_size=32)

    # Evaluate the model on the validation set
    _, accuracy = model.evaluate(X_val, y_val)
    accuracies.append(accuracy)

# Calculate the mean and standard deviation of accuracies
mean_accuracy = np.mean(accuracies)
std_accuracy = np.std(accuracies)
print(f"\nMean Accuracy: {mean_accuracy:.4f}")
print(f"Standard Deviation: {std_accuracy:.4f}")

# Save the model
model.save('my_model.keras')

Keep getting an error: Traceback (most recent call last): File "model.py", line 44, in image = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE)

I have checked many times that the file path is correct, and I was running the model previously on Google Colab but due to being too slow I moved over to training my model on my universities computer cluster. I have no idea why it suddenly doesnt work.

welcome. complete traceback required. [mre]. insert this line: `assert os.path.exists(image_path), os.getcwd()` -- and what is the version of opencv-python you are using? — Christoph Rackwitz, Jul 26 '23 at 20:46

score 0 · Answer 1 · answered Jul 27 '23 at 12:04

Without the full traceback it's a bit hard to guess, but I think here is your solution.
The / in train_folder_path = os.path.join(data_path, "/train") kills everything in data_path and train_folder_path should be now /train.
As Christoph Rackwitz mentioned in the comments, a full traceback/better error message is really helpful, you cut out almost the complete traceback message which would most likely include hints on the type of error. Also debbuging and testing with print(path) can be helpful.

Issue reading the image using cv2.imread() when training my model

1 Answers1