先放上源码

import os
import shutil
import subprocess
import tarfileimport os
import pandas as pd
import numpy as np
from PIL import Image# Set up Kaggle API credentials
kaggle_json_path = os.path.expanduser("~/.kaggle/kaggle.json")if not os.path.isfile(kaggle_json_path):username = input("Kaggle username: ")api_key = input("Kaggle API key: ")# Create the directory if it doesn't existos.makedirs(os.path.dirname(kaggle_json_path), exist_ok=True)# Write the credentials to the kaggle.json filewith open(kaggle_json_path, "w") as file:file.write(f'{{"username":"{username}","key":"{api_key}"}}')# Set file permissions to read and write for the owner onlyos.chmod(kaggle_json_path, 0o600)# Importing kaggle will authenticate automatically
import kaggle# Command to authenticate and download the dataset
api_command = "kaggle competitions download -c challenges-in-representation-learning-facial-expression-recognition-challenge -f fer2013.tar.gz"# Execute the command
try:subprocess.run(api_command, shell=True, check=True)
except subprocess.CalledProcessError as e:print("An error occurred while downloading the dataset. Please double-check your Kaggle API key.")os.remove(kaggle_json_path)print("Preparing dataset..")
with tarfile.open("fer2013.tar.gz", "r") as tar:tar.extractall("fer2013")output_folder_path = "FER2013"# Load the CSV file into a Pandas DataFrame
df = pd.read_csv("fer2013/fer2013/fer2013.csv")# Define a dictionary to map emotion codes to labels
emotion_labels = {"0": "Angry","1": "Disgust","2": "Fear","3": "Happy","4": "Sad","5": "Surprise","6": "Neutral",
}# Create the output folders and subfolders if they do not exist
if not os.path.exists(output_folder_path):os.makedirs(output_folder_path)
for usage in ["train", "val", "test"]:usage_folder_path = os.path.join(output_folder_path, usage)if not os.path.exists(usage_folder_path):os.makedirs(usage_folder_path)for label in emotion_labels.values():subfolder_path = os.path.join(usage_folder_path, label)if not os.path.exists(subfolder_path):os.makedirs(subfolder_path)# Loop over each row in the DataFrame
for index, row in df.iterrows():# Extract the image data from the rowpixels = row["pixels"].split()img_data = [int(pixel) for pixel in pixels]img_array = np.array(img_data).reshape(48, 48)img = Image.fromarray(img_array.astype("uint8"), "L")# Get the emotion label and determine the output subfolder based on the Usage columnemotion_label = emotion_labels[str(row["emotion"])]if row["Usage"] == "Training":output_subfolder_path = os.path.join(output_folder_path, "train", emotion_label)elif row["Usage"] == "PublicTest":output_subfolder_path = os.path.join(output_folder_path, "val", emotion_label)else:output_subfolder_path = os.path.join(output_folder_path, "test", emotion_label)# Save the image to the output subfolderoutput_file_path = os.path.join(output_subfolder_path, f"{index}.jpg")img.save(output_file_path)print("Deleting temporary files..")
os.remove("fer2013.tar.gz")
shutil.rmtree("fer2013")