AuraBloom.py

import streamlit as st
from PIL import Image
import numpy as np
from joblib import load
from skimage.transform import resize
import torch
import os
import sys

# Ensure to run these commands in your terminal first:
# pip install git+https://github.com/FacePerceiver/facer.git@main
# pip install timm
# git clone https://github.com/FacePerceiver/facer.git

# Set the path for the 'facer' module
sys.path.append('facer')

import facer

# Load face parsing model
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
face_detector = facer.face_detector('retinaface/mobilenet', device=device)
face_parser = facer.face_parser('farl/lapa/448', device=device)

# Define the monk scale colors
monk_scale = {
    'Class2': (243, 231, 219),  # f3e7db
    'Class3': (247, 234, 208),  # f7ead0
    'Class4': (234, 218, 186),  # eadaba
    'Class5': (215, 189, 150),  # d7bd96
    'Class6': (160, 126, 86),   # a07e56
    'Class7': (130, 92, 67),    # 825c43
    'Class8': (96, 65, 52),     # 604134
    'Class9': (58, 49, 42),     # 3a312a
    'Class10': (41, 36, 32),    # 292420
}

# Function to convert RGB tuple to hex color code
def rgb_to_hex(rgb):
    return '#{:02x}{:02x}{:02x}'.format(*rgb)

# Mapping of Monk classes to colors using monk_scale
monk_colors = {
    '1': [rgb_to_hex(monk_scale['Class2']), rgb_to_hex(monk_scale['Class3']), rgb_to_hex(monk_scale['Class4'])],
    '2': [rgb_to_hex(monk_scale['Class5']), rgb_to_hex(monk_scale['Class6'])],
    '3': [rgb_to_hex(monk_scale['Class7']), rgb_to_hex(monk_scale['Class8'])],
    '4': [rgb_to_hex(monk_scale['Class9']), rgb_to_hex(monk_scale['Class10'])],
    'default': '#808080'  # Default color for unexpected classes
}

# Mapping of model's output classes to monk classes
class_mapping = {
    0: '1',  # Map model class 0 to monk class 1
    1: '2',  # Map model class 1 to monk class 2
    2: '3',  # Map model class 2 to monk class 3
    3: '4',  # Map model class 3 to monk class 4
    # Add more mappings if needed
}

# Function to load the model
def load_model():
    model_path = r"C:\Users\ramam\svm_model3.joblib"  # Adjust the path to your model
    model = load(model_path)
    return model

# Function to parse face and extract skin region
def parse_face(image):
    # Ensure the image has 3 channels (RGB)
    if image.mode != 'RGB':
        image = image.convert('RGB')
    
    image_data = np.array(image)
    
    # Check if the image has 3 channels
    if image_data.shape[2] != 3:
        raise ValueError("Image does not have 3 channels (RGB).")
    
    image_tensor = torch.from_numpy(image_data.astype('float32')).permute(2, 0, 1).unsqueeze(0).to(device)
    faces = face_detector(image_tensor)

    if faces:
        parsed_faces = face_parser(image_tensor, faces)
        if 'seg' in parsed_faces:
            seg_logits = parsed_faces['seg']['logits']
            seg_probs = torch.sigmoid(seg_logits)
            binary_mask = seg_probs[0, 1, :, :] > 0.5
            binary_mask = binary_mask.cpu().numpy()
            binary_mask_3d = np.repeat(binary_mask[:, :, np.newaxis], 3, axis=2)
            skin_region = image_data * binary_mask_3d
            return skin_region.astype(np.uint8)
    return None

# Function to make predictions
def classify_image(image, model):
    parsed_image = parse_face(image)
    if parsed_image is not None:
        image_resized = resize(parsed_image, (128, 128), anti_aliasing=True)  # Resize to 128x128
        image_reshaped = image_resized.reshape(1, -1)  # Reshape to match the model input
        if image_reshaped.shape[1] == 49152:  # Check if resizing is correct
            image_padded = np.pad(image_reshaped, ((0, 0), (0, 65536 - 49152)), 'constant')
        else:
            raise ValueError("Unexpected number of features after reshaping.")
        prediction = model.predict(image_padded)
        return prediction[0], parsed_image
    else:
        raise ValueError("Face parsing failed.")

# Load the model
model = load_model()

# Function to display the Monk class color
def display_monk_class_color(prediction):
    st.write(f"Prediction: {prediction}")  # Debugging
    monk_class = class_mapping.get(prediction, 'default')
    colors = monk_colors.get(monk_class, monk_colors['default'])  # Default to gray if class not found
    st.write(f"Monk Class: {monk_class}")
    for color in colors:
        st.markdown(f"<div style='width:100px; height:50px; background-color:{color};'></div>", unsafe_allow_html=True)

# Streamlit app
st.title('Skin Tone Classification')

uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])
if uploaded_file is not None:
    image = Image.open(uploaded_file)
    st.image(image, caption='Uploaded Image.', use_column_width=True)

    if st.button('Classify'):
        try:
            prediction, parsed_image = classify_image(image, model)
            display_monk_class_color(prediction)
            st.image(parsed_image, caption='Parsed Image.', use_column_width=True)
        except ValueError as e:
            st.error(f"Error: {e}")