图像加载函数实现报错及结果不符问题求助

Problem Description #

I need to implement a load_data function that reads 3 plate images (corresponding to B, G, R channels, top to bottom) from a directory and returns a list of these images. The plates directory is located alongside my Notebook in the "Deep Learning in Computer Vision" folder.

My first attempt at the code threw a FileNotFoundError saying it couldn't find the directory 'C':

def load_data(dir_name ='C:/Users/ASUS/Desktop/Self_Learning/Coursera/Deep Learning in Computer Vision/plates'):
    im_list=[]
    for i in dir_name:
        im=np.load(i)
        im_list.append(im)
    return im_list
pass
plates = load_data()

I realized this was because I was iterating over each character in the directory path string instead of the files inside the directory.

My second attempt didn't work either—it returned a list of path characters instead of the image data I expected (I was hoping for a list of loaded image arrays, but got a mess of individual characters from file paths):

import os
def load_data(dir_name ='C:/Users/ASUS/Desktop/Self_Learning/Coursera/Deep Learning in Computer Vision/plates'):
    im_list=[]
    for f in os.listdir(dir_name):
        fpath = os.path.join(dir_name, f)
        for im in fpath:
            im_list.append(im)
    print(im_list)
plates = load_data()

I need help fixing this function to correctly load the images into a usable list.

Solution #

The core issue in both attempts was incorrect iteration: you were looping over strings (the directory path or file path) instead of loading the actual image files. Here's the corrected function that properly loads your numpy image files:

import os
import numpy as np

def load_data(dir_name ='C:/Users/ASUS/Desktop/Self_Learning/Coursera/Deep Learning in Computer Vision/plates'):
    im_list = []
    # Get sorted list of files to ensure consistent B/G/R channel order
    sorted_files = sorted(os.listdir(dir_name))
    
    for filename in sorted_files:
        # Skip non-numpy files (adjust extension if your images use a different format)
        if filename.endswith('.npy'):
            full_file_path = os.path.join(dir_name, filename)
            # Load the image file and add it to the list
            image = np.load(full_file_path)
            im_list.append(image)
    
    return im_list

# Load the plate images
plates = load_data()
# Verify the result
print(f"Successfully loaded {len(plates)} images")

Key Fixes Explained: #

• Correct File Iteration: We loop over the filenames returned by os.listdir(dir_name) instead of iterating over characters in the path string.
• Valid Path Construction: os.path.join(dir_name, filename) creates a cross-platform compatible full path to each image, avoiding path errors.
• Consistent Order: sorted(os.listdir(dir_name)) ensures we load the images in a predictable sequence (critical for maintaining B/G/R channel alignment).
• Filter Valid Files: The endswith('.npy') check skips any unrelated files in the directory (like hidden system files) that could cause loading errors.
• Proper Image Loading: We use np.load(full_file_path) to actually load the numpy image data and append it to the list, instead of adding path characters.

This will return a list of your 3 channel images, ready for further computer vision processing.

内容的提问来源于stack exchange，提问作者Rim Sleimi