danhtran2mind's picture
Update README.md
70e8f53 verified
|
raw
history blame
3.08 kB
metadata
library_name: keras
license: mit
language:
  - en
pipeline_tag: image-to-image

Metrics

PSNR

  • Validation set: 21.70

Usage

Download Model

git clone https://huggingface.co/danhtran2mind/autoencoder-grayscale2color-landscape
cd autoencoder-grayscale2color-landscape
git lfs pull

Import Libraries

from PIL import Image
import os
import numpy as np
import tensorflow as tf
import requests
from skimage.color import lab2rgb

from models.auto_encoder_gray2color import SpatialAttention

Load Model file

# Load the saved model once at startup
load_model_path = "./ckpts/best_model.h5"

print(f"Loading model from {load_model_path}...")
loaded_autoencoder = tf.keras.models.load_model(
    load_model_path,
    custom_objects={'SpatialAttention': SpatialAttention}
)

Inferenc Step

# Assuming process_image function is defined as provided
# and required imports (tensorflow, skimage.color for lab2rgb) are available

def display_images_pil(input_path):
    # Read input image
    input_img = Image.open(input_path)
    
    # Process the image using the provided function
    output_img = process_image(input_img)
    
    # Display input image
    input_img.show(title="Input Image")
    
    # Display output image
    output_img.show(title="Output Image")

def process_image(input_img):
    # Store original input dimensions
    original_width, original_height = input_img.size

    # Convert PIL Image to grayscale and resize to model input size
    img = input_img.convert("L")  # Convert to grayscale (single channel)
    img = img.resize((WIDTH, HEIGHT))  # Resize to 512x512 for model
    img_array = tf.keras.preprocessing.image.img_to_array(img) / 255.0  # Normalize to [0, 1]
    img_array = img_array[None, ..., 0:1]  # Add batch dimension, shape: (1, 512, 512, 1)

    # Run inference (assuming loaded_autoencoder predicts a*b* channels)
    output_array = loaded_autoencoder.predict(img_array)  # Shape: (1, 512, 512, 2) for a*b*
    print("output_array shape: ", output_array.shape)

    # Extract L* (grayscale input) and a*b* (model output)
    L_channel = img_array[0, :, :, 0] * 100.0  # Denormalize L* to [0, 100]
    ab_channels = output_array[0] * 128.0  # Denormalize a*b* to [-128, 128]

    # Combine L*, a*, b* into a 3-channel L*a*b* image
    lab_image = np.stack([L_channel, ab_channels[:, :, 0], ab_channels[:, :, 1]], axis=-1)  # Shape: (512, 512, 3)

    # Convert L*a*b* to RGB
    rgb_array = lab2rgb(lab_image)  # Convert to RGB, output in [0, 1]
    rgb_array = np.clip(rgb_array, 0, 1) * 255.0  # Scale to [0, 255]
    rgb_image = Image.fromarray(rgb_array.astype(np.uint8), mode="RGB")  # Create RGB PIL image

    # Resize output image to match input image resolution
    rgb_image = rgb_image.resize((original_width, original_height), Image.Resampling.LANCZOS)

    return rgb_image


# Example usage
if __name__ == "__main__":
    # Replace 'input_image.jpg' with the path to your image
    image_path = "<input_image.jpg>"
    display_images_pil(image_path)