my_bp_dataset.py

import os
import pandas as pd
from PIL import Image
from datasets import Dataset, DatasetInfo, Features, Value, Image as HFImage
import datasets

def load_bp_dataset(csv_path, image_folder):
    # Load CSV file as a DataFrame
    df = pd.read_csv(csv_path)
    
    # Create a function to load images based on the paths in the CSV
    def load_image(image_path):
        return Image.open(os.path.join(image_folder, image_path))

    # Add image column to the DataFrame by loading images using paths in 'Image_Path'
    df['image'] = df['Image_Path'].apply(load_image)

    # Convert the DataFrame into a Hugging Face Dataset
    dataset = Dataset.from_pandas(df)
    return dataset

if __name__ == "__main__":
    # Provide paths to CSV and image folder
    csv_path = '/mnt/data/image_metadata.csv'  # Path to your CSV file
    image_folder = '/mnt/data/BP_images'  # The folder where images are stored
    
    # Load dataset
    bp_dataset = load_bp_dataset(csv_path, image_folder)

    # Display an example to verify everything is loaded correctly
    print(bp_dataset[0])

    # Optional: Save the dataset to Hugging Face format if needed
    bp_dataset.push_to_hub("your_username/your_dataset_name")