Generate CSV file to actual image

[tien-m-duong]

Dear Helen,

Thank you for this interesting work. I was able to the generate a CSV file using GCSbc models. I wonder if there is a way to convert the .csv file to the actual image. I tried to search around but don't see any options. Thank you so much.

Best,
Tien

[helenmand]

Tien,
the GaussianCopulaSynthesizer is used for tabular data. In order to use it in images you must have converted somehow your images to a CSV. For example:

image = Image.open(image_path)

# convert image to numpy array 
image_array = np.array(image)

# flatten into a 1D vector 
flattened_image = image_array.flatten()

Now after using this data, in GaussianCopulaSynthesizer, you can use the generated CSV file to get the images like this:

df = pd.read_csv(generated_csv_path)
image_data = np.array(df.iloc[0]) # for the single image

# reshape the data to its original shape 
image_shape = (564, 564, 3)  # your actual image size
image_array = image_data.reshape(image_shape)

# convert the numpy array back to an image
image = Image.fromarray(image_array.astype(np.uint8))

I hope this helped.

[tien-m-duong]

Thank you for your explanation. Just want to confirm my understanding. The problem will run as:

real image .png file(let's say chest X-ray) => convert to numpy array through your first code => run through GaussianCopulaSynthesizer => generated csv file => run through second code to get synthetic chest X-ray images

Please let me know if my understanding is incorrect. Thank you so much