CellCTC/mask_convert.py

import os
import numpy as np
from PIL import Image
import random
import matplotlib.pyplot as plt
from matplotlib.colors import ListedColormap

# Base directory for the datasets
base_dir = "/l/users/sahal.mullappilly/Komal/documents/Cell/CTCdataset"

# Function to generate unique colors for each cell ID
def generate_colormap():
    # Define a colormap with enough unique colors
    colors = plt.cm.get_cmap('tab20', 20)  # You can adjust the number of colors here
    color_list = [colors(i) for i in range(colors.N)]
    # Generate a list of unique RGB colors for each unique cell ID
    return ListedColormap(color_list)

# Function to convert ground truth tif to colored png
def convert_gt_to_colored_png(gt_tif_path, output_png_path, colormap):
    try:
        # Open the ground truth tiff image using PIL
        gt_img = Image.open(gt_tif_path)

        # Convert to a numpy array
        gt_array = np.array(gt_img)

        # Create an empty colored image using the colormap
        colored_gt = np.zeros((gt_array.shape[0], gt_array.shape[1], 3), dtype=np.uint8)

        # For each unique cell ID, assign a unique color
        unique_ids = np.unique(gt_array)
        for cell_id in unique_ids:
            if cell_id == 0:  # Assuming 0 represents background, no need to color it
                continue
            color = colormap(cell_id % 20)[:3]  # Normalize cell ID to map to the colormap
            # Assign the color to all pixels with this cell ID
            colored_gt[gt_array == cell_id] = np.array(color) * 255

        # Convert the array back to an image
        colored_gt_img = Image.fromarray(colored_gt)
        # Save the colored image as PNG
        colored_gt_img.save(output_png_path)
        print(f"Converted {gt_tif_path} to {output_png_path}")
    except Exception as e:
        print(f"Error converting {gt_tif_path}: {e}")

# List of datasets and their corresponding subfolders (01, 02)
datasets = [
    # {"dataset_name": "Fluo-C2DL-Huh7", "subfolders": ["01", "02"]},
    # {"dataset_name": "Fluo-C2DL-MSC", "subfolders": ["01", "02"]},
    # {"dataset_name": "Fluo-N2DH-SIM+", "subfolders": ["01", "02"]},
    # {"dataset_name": "Fluo-N2DL-HeLa", "subfolders": ["01", "02"]},
    {"dataset_name": "Fluo-N2DH-GOWT1", "subfolders": ["01", "02"]},
#     {"dataset_name": "PhC-C2DL-PSC", "subfolders": ["01", "02"]},
#     {"dataset_name": "DIC-C2DH-HeLa", "subfolders": ["01", "02"]},
#     {"dataset_name": "BF-C2DL-HSC", "subfolders": ["01", "02"]},
#     {"dataset_name": "BF-C2DL-MuSC", "subfolders": ["01", "02"]},
#     {"dataset_name": "PhC-C2DH-U373", "subfolders": ["01", "02"]}
]

# Path for the main "Annotations" directory
main_annotations_dir = os.path.join(base_dir, "Annotations")
os.makedirs(main_annotations_dir, exist_ok=True)

# Create a colormap for unique cell IDs
colormap = generate_colormap()

# Process each dataset and its subfolders
for dataset in datasets:
    for subfolder in dataset["subfolders"]:
        # Paths for the input ground truth and the output directory
        input_folder = os.path.join(base_dir, "datasets", dataset["dataset_name"], f"{subfolder}_GT", "TRA")
        output_folder = os.path.join(main_annotations_dir, f"{dataset['dataset_name']}-{subfolder}")
        
        # Create output directory for each subfolder inside "Annotations"
        os.makedirs(output_folder, exist_ok=True)

        # Process each .tif file in the subfolder
        for root, dirs, files in os.walk(input_folder):
            for file in files:
                if file.endswith(".tif"):
                    gt_tif_path = os.path.join(root, file)
                    # Define the path for the new png file
                    png_filename = file.replace(".tif", ".png")
                    png_path = os.path.join(output_folder, png_filename)
                    
                    # Convert the ground truth .tif to colored .png
                    convert_gt_to_colored_png(gt_tif_path, png_path, colormap)

print("Ground truth TIFF images converted to colored PNG in the 'Annotations' folder!")