Hey there, let's break down why you're hitting this error and fix it step by step!

What's Causing the Error? #

The error pops up because cv2.imread(data[i]) is returning None—meaning OpenCV couldn't successfully load the image file you're trying to read. When you try to access the .shape attribute of a None object, Python throws that AttributeError.

Common reasons this happens:

• The image path is incorrect, or the file doesn't exist (maybe glob picked up a non-image file?)
• The image file itself is corrupted and can't be loaded
• The image format isn't supported by OpenCV (rare for common formats like JPG/PNG, but possible for niche ones)

How to Fix It #

We'll modify your code to add checks for successful image loading, plus make path handling more cross-platform friendly. Here's the updated code:

import os
import glob
import cv2

# Assume your classes, TRAIN_DIR, and class2label variables are already defined
categories = [] 
image_name = [] 
image_id = [] 
image_format = [] 
labels = [] 
img_path = [] 
img_height = [] 
img_width = [] 

for cat in classes: 
    path = os.path.join(TRAIN_DIR, cat, "*") 
    data = glob.glob(path) 
    for img_file in data:
        # First, attempt to load the image and check if it succeeded
        img = cv2.imread(img_file)
        if img is None:
            print(f"Warning: Failed to load image at {img_file}. Skipping this file.")
            continue
        
        # Now safely get the image dimensions
        h, w, c = img.shape 
        
        # Use os.path utilities for cross-platform path handling (works on Windows/macOS/Linux)
        cls = os.path.basename(os.path.dirname(img_file))
        img_name = os.path.basename(img_file)
        img_id_part, img_format_part = os.path.splitext(img_name)
        # Remove the leading dot from the format (e.g., ".jpg" becomes "jpg")
        img_format_part = img_format_part.lstrip('.')
        
        # Append data to your lists
        img_height.append(h) 
        img_width.append(w) 
        categories.append(cls) 
        image_name.append(img_name) 
        image_id.append(img_id_part) 
        image_format.append(img_format_part) 
        labels.append(class2label[cls]) 
        img_path.append(img_file)

Key Improvements in the Updated Code #

• Image loading check: We added if img is None to catch failed loads, print a warning, and skip the problematic file instead of crashing.
• Cross-platform path handling: Replaced manual / splitting with os.path methods, which work regardless of your operating system's path separator.
• Cleaner format extraction: Used os.path.splitext to safely get the file extension, then stripped the leading dot for consistency.

Extra Troubleshooting Tips #

• Double-check that TRAIN_DIR points to the correct root directory of your training images.
• Open the files mentioned in the warning messages to see if they're corrupted or not actually image files.
• If glob is picking up non-image files, narrow your path pattern to specific formats (e.g., "*.jpg" or "*.png") instead of "*".

内容的提问来源于stack exchange，提问作者farhad_bme