app.py

import gradio as gr
from astropy.io import fits
import matplotlib.pyplot as plt
import numpy as np
import io
from PIL import Image
import astropy.units as u
from astropy.wcs import WCS
from astropy.coordinates import SkyCoord
from astropy import coordinates as coord
from astropy.wcs.utils import skycoord_to_pixel
from astroquery.simbad import Simbad
import pandas as pd
import matplotlib.patches as patches
# Increase the limit (set to a value larger than the pixel count of your image)
Image.MAX_IMAGE_PIXELS = None
plt.style.use('dark_background')

# Initialize globals
global_dataframe = pd.DataFrame()
global_data = None
global_header = None

def show_csv(file):
    """
    Displays the uploaded CSV file as a table.
    """
    global global_dataframe
    
    try:
        # Read the CSV file into a pandas DataFrame
        df = pd.read_csv(file.name, index_col=0)
        global_dataframe = df  # Store the dataframe globally for filtering
        # Extract unique types from the "type" column
        if "TYPE" in df.columns:
            unique_types = df["TYPE"].unique().tolist()
            return df, gr.CheckboxGroup(label="Select Catalogue", choices=unique_types, value=unique_types, interactive=True)
        else:
            return "Error: CSV does not contain a 'type' column.", None
    except Exception as e:
        return f"Error: {str(e)}", None
    
# Define a function to be called when the button is clicked
def query_update_table():
    """
    Displays the uploaded CSV file as a table.
    """
    global global_dataframe, global_header, global_data
    
    try:
        # Read the CSV file into a pandas DataFrame
        #df = pd.read_csv('dataframe.csv', index_col=0)

        Simbad.TIMEOUT = 120

        # Define the specific coordinates
        wcs = WCS(global_header).dropaxis(2)
        center_ra = global_header['CRVAL1']
        center_dec = global_header['CRVAL2']
        target_coord = SkyCoord(ra=center_ra, dec=center_dec, unit=(u.deg, u.deg), frame='icrs')
        print(center_ra, center_dec)
        # define the search radius
        radius_deg = max([abs(global_header['CDELT1']),abs(global_header['CDELT2'])])*max([global_header['NAXIS1'],global_header['NAXIS2']])
        radius_deg *= 1
        # Set up the query criteria
        if target_coord.dec.deg > 0:
            custom_query = f"region(CIRCLE, {target_coord.ra.deg} +{target_coord.dec.deg}, {radius_deg}d)"
        else:
            custom_query = f"region(CIRCLE, {target_coord.ra.deg} {target_coord.dec.deg}, {radius_deg}d)"
        
        print(f'Query={custom_query}')

        result_table = Simbad.query_criteria(custom_query, otype='galaxy')
        
        print("received feedback from simbad!!!")
        print(result_table)
        df = result_table.to_pandas().set_index('main_id')
        print(df.columns)
        df['Pixel_Position'] = [skycoord_to_pixel(SkyCoord(v[0],v[1], unit=(u.deg, u.deg), frame='icrs'), wcs) for v in df[['ra','dec']].values]
        print(df['Pixel_Position'])
        df['px'] = df['Pixel_Position'].apply(lambda x: int(x[0]))
        df['py'] = df['Pixel_Position'].apply(lambda x: int(x[1]))
        mask = (df.px>0)&(df.px< global_data.shape[1])&(df.py>0)&(df.py<global_data.shape[0])
        print(df)
        df = df[mask]
        df = df.reset_index()
        df['TYPE'] = df['main_id'].apply(lambda x: x.split(' ')[0].split('+')[0])
        df = df.sort_values(by=['px', 'py'], ascending=[True, True]).reset_index(drop=True)
        print(df)
        df = df.iloc[:200]
        global_dataframe = df  # Store the dataframe globally for filtering

        # Extract unique types from the "type" column
        if "TYPE" in df.columns:
            unique_types = df["TYPE"].unique().tolist()
            return df, gr.CheckboxGroup(label="Select Catalogue", choices=unique_types, value=unique_types, interactive=True)
        else:
            return "Error: CSV does not contain a 'type' column.", None
    except Exception as e:
        return f"Error: {str(e)}", None


def load_fits_image(file, type_checkboxes, title, axis_options, num_rows, patch_size, fontsize, alpha, linewidth, scale, patch_color, sort_method):
    """
    Displays the data from the uploaded FITS file.
    """
    global global_header, global_data

    # Open the FITS file
    hdu = fits.open(file)
    data = hdu[0].data  # Access the primary HDU data
    data = np.swapaxes(np.swapaxes(data,0,2),0,1)#.astype(np.float)
    #data = (data*255).astype(np.uint8)  # Access the primary HDU data
    global_data = data

    # get fits header
    header = hdu[0].header
    global_header = header
 #selected_types, title, selected_axis_options, num_rows, patch_size, patch_color, sort_method
    return update_images_and_tables(type_checkboxes, title, axis_options, num_rows, patch_size, fontsize, alpha, linewidth, scale, patch_color, sort_method)
 
def update_images_and_tables(selected_types, title, selected_axis_options, num_rows, patch_size, fontsize, alpha, linewidth, scale, patch_color, sort_method):
    global global_dataframe, global_header, global_data
    
    if selected_types and not global_dataframe.empty:
        # Filter the dataframe based on the selected types
        filtered_df = global_dataframe[global_dataframe["TYPE"].isin(selected_types)]
        mask = (filtered_df.px-patch_size//2 > 0)&(filtered_df.px+patch_size//2 < global_data.shape[1])&(filtered_df.py-patch_size//2 > 0)&(filtered_df.py+patch_size//2 < global_data.shape[0])
        filtered_df = filtered_df[mask]
    else:
        filtered_df = None

    if not filtered_df is None:
        # Sort the dataframe based on the sorting method
        if sort_method == "by Catalogue":
            filtered_df = filtered_df.sort_values(by=['px', 'py'], ascending=[True, True])
            filtered_df = filtered_df.sort_values(by='TYPE', ascending=True).reset_index(drop=True)
        elif sort_method == "by x":
            filtered_df = filtered_df.sort_values(by=['px', 'py'], ascending=[True, True]).reset_index(drop=True)
        elif sort_method == "by y":
            filtered_df = filtered_df.sort_values(by=['py', 'px'], ascending=[True, True]).reset_index(drop=True)

    try:
        
        wcs = WCS(global_header).dropaxis(2)
        
        ratio = global_data.shape[0]/global_data.shape[1]
        # Plot WCS
        fig = plt.figure(figsize=(ratio*scale,scale))
        ax = fig.add_subplot(projection=wcs, label='overlays')
        ax.imshow(global_data, origin='lower')

        #if not filtered_df is None:
        #    filtered_df.plot.scatter(x='px', y='py', ax=ax, s=15, c=patch_color)

        if "with Grid" in selected_axis_options:
            ax.coords.grid(True, color='white', ls='-', alpha=.5)
        if "with Axis Annotation" in selected_axis_options:
            ax.coords[0].set_axislabel('Right Ascension (J2000)', fontsize=fontsize+2)
            ax.coords[1].set_axislabel('Declination (J2000)', fontsize=fontsize+2)
        else:
            ax.axis('off')
        plt.title(title, fontsize=fontsize+4)

        if not filtered_df is None:
            all_patches = []
            for i,row in filtered_df.iterrows():
                rect = patches.Rectangle((row.px-patch_size//2, row.py-patch_size//2), patch_size, patch_size, alpha=alpha, linewidth=linewidth, edgecolor=patch_color, facecolor='none')
                ax.add_patch(rect)
                ax.text(row.px,row.py+patch_size//2,str(i+1),
                        ha='center',va='bottom',color=patch_color,fontsize=fontsize)
                patch = global_data[row.py-patch_size//2:row.py+patch_size//2,row.px-patch_size//2:row.px+patch_size//2]
                all_patches.append(patch)
        plt.tight_layout()
        # Convert the plot to an image
        buf = io.BytesIO()
        plt.savefig(buf, format='png', bbox_inches='tight', pad_inches=.1, dpi=200)
        plt.close(fig)
        buf.seek(0)
        # Convert buffer to PIL Image
        image = Image.open(buf)

        if not filtered_df is None:
            
            m = num_rows
            n = int(np.ceil(len(filtered_df)/m))

            second_scale=max([1,scale//3])
            fig, axarr = plt.subplots(n,m,figsize=(m*second_scale,n*second_scale))
            for i, row in filtered_df.iterrows():
                    ax = axarr[i//m,i%m]
                    ax.imshow(all_patches[i][::-1])
                    ax.set_title(row.main_id, fontsize=fontsize-2)
                    ax.set_xticks([])
                    ax.set_yticks([])
                    ax.text(2,2,str(i+1)[:30],ha='left',va='top',fontsize=fontsize+6)
            for i in np.arange(len(all_patches),m*n):
                    ax = axarr[i//m,i%m]
                    ax.axis('off')
            plt.tight_layout()

            # Convert the plot to an image
            second_buf = io.BytesIO()
            plt.savefig(second_buf, format='png', bbox_inches='tight', pad_inches=.1, dpi=200)
            plt.close(fig)
            second_buf.seek(0)
            # Convert buffer to PIL Image
            patches_image = Image.open(second_buf)

            return filtered_df, image, patches_image
        else:
            return filtered_df, image, None
    
    except Exception as e:
        return f"Error: {str(e)}"


# Gradio interface
with gr.Blocks(css=".btn-green {background-color: green; color: white;}") as gui:
    gr.Markdown("# What's in my image?")
    # Options Area
    with gr.Row() as options_gui:
        num_rows = gr.Number(label="Number of Rows", value=16, minimum=2, precision=0, interactive=True)
        title = gr.Textbox(label="Image Title", value="Custom Title", interactive=True)
        patch_size = gr.Slider(label="Patch Size", minimum=16, maximum=128, step=8, value=32, 
        interactive=True)
        fontsize = gr.Slider(label="Fontsize", minimum=6, maximum=26, step=1, value=10, 
        interactive=True)
        alpha = gr.Slider(label="Alpha", minimum=0., maximum=1., step=.1, value=1., 
        interactive=True)
        linewidth = gr.Slider(label="Linewidth", minimum=1, maximum=4, step=1, value=1, 
        interactive=True)
        scale = gr.Slider(label="Scale", minimum=1, maximum=20, step=1, value=8, 
        interactive=True)

        patch_color = gr.ColorPicker(label="Patch Color", value="#FFFFFF", interactive=True)
        sort_method = gr.Dropdown(label="Sorting Method", choices=["by Catalogue", "by x", "by y"], value="by Catalogue", interactive=True)
        axis_options = gr.CheckboxGroup(
            label="Select options",
            choices=["with Grid", "with Axis Annotation"],
            value=["with Grid", "with Axis Annotation"],  # Preselected values
            interactive=True  # Makes it interactive
        )
    gr.Markdown("Upload a plate solved `.fits` file (32 bit) to display its content.")
    
    file_input = gr.File(label="Upload .fits File", type="filepath")
    #file_input_csv = gr.File(label="Upload .csv File")
    greet_button = gr.Button("Query Simbad for Galaxies")  # Create the button
    fits_image = gr.Image(label="Input Image", type="pil")
    type_checkboxes = gr.CheckboxGroup(label="Select Catalogue")
    patches_image = gr.Image(label="Patches Image", type="pil")
    csv_table = gr.DataFrame(label="CSV Table")

    track_options = [type_checkboxes, title, axis_options, num_rows, patch_size, fontsize, alpha, linewidth, scale, patch_color, sort_method]

    file_input.change(load_fits_image, 
                      inputs=[file_input] + track_options, 
                      outputs=[csv_table,fits_image,patches_image])
        
    for option_i in track_options:
        option_i.change(update_images_and_tables,
                       inputs=track_options, 
                      outputs=[csv_table,fits_image,patches_image])

    # Display CSV table
    #file_input_csv.change(show_csv, 
    #                      inputs=file_input_csv, 
    #                      outputs=[csv_table, type_checkboxes])
        
    greet_button.click(query_update_table, inputs=None, outputs=[csv_table, type_checkboxes])

    # Update the selected checkboxes change
    type_checkboxes.change(update_images_and_tables, 
                           inputs=track_options,
                           outputs=[csv_table,fits_image,patches_image])

gui.launch(debug=True)