CivitAI / app.py
bgaspra's picture
Update app.py
64e976b verified
import os
import requests
from tqdm import tqdm
from datasets import load_dataset
import numpy as np
from tensorflow.keras.applications.resnet50 import ResNet50, preprocess_input
from tensorflow.keras.preprocessing import image
from sklearn.neighbors import NearestNeighbors
import joblib
from PIL import UnidentifiedImageError, Image
import gradio as gr
# Load the dataset
dataset = load_dataset("thefcraft/civitai-stable-diffusion-337k")
# Take a subset of the dataset
subset_size = 2700 # Reduce the subset size for faster execution in Spaces
dataset_subset = dataset['train'].shuffle(seed=42).select(range(subset_size))
# Directory to save images
image_dir = 'civitai_images'
os.makedirs(image_dir, exist_ok=True)
# Load the ResNet50 model pretrained on ImageNet
model = ResNet50(weights='imagenet', include_top=False, pooling='avg')
# Function to extract features
def extract_features(img_path, model):
img = image.load_img(img_path, target_size=(224, 224))
img_array = image.img_to_array(img)
img_array = np.expand_dims(img_array, axis=0)
img_array = preprocess_input(img_array)
features = model.predict(img_array)
return features.flatten()
# Extract features for a sample of images
features = []
image_paths = []
model_names = []
for sample in tqdm(dataset_subset):
img_url = sample['url'] # Adjust based on the correct column name
model_name = sample['Model'] # Adjust based on the correct column name
img_path = os.path.join(image_dir, os.path.basename(img_url))
# Download the image
try:
response = requests.get(img_url)
response.raise_for_status() # Check if the download was successful
if 'image' not in response.headers['Content-Type']:
raise ValueError("URL does not contain an image")
with open(img_path, 'wb') as f:
f.write(response.content)
# Extract features
try:
img_features = extract_features(img_path, model)
features.append(img_features)
image_paths.append(img_path)
model_names.append(model_name)
except UnidentifiedImageError:
print(f"UnidentifiedImageError: Skipping file {img_path}")
os.remove(img_path)
except requests.exceptions.RequestException as e:
print(f"RequestException: Failed to download {img_url} - {e}")
# Convert features to numpy array
features = np.array(features)
# Build the NearestNeighbors model
nbrs = NearestNeighbors(n_neighbors=5, algorithm='ball_tree').fit(features)
# Save the model and features
joblib.dump(nbrs, 'nearest_neighbors_model.pkl')
np.save('image_features.npy', features)
np.save('image_paths.npy', image_paths)
np.save('model_names.npy', model_names)
# Load the NearestNeighbors model and features
nbrs = joblib.load('nearest_neighbors_model.pkl')
features = np.load('image_features.npy')
image_paths = np.load('image_paths.npy', allow_pickle=True)
model_names = np.load('model_names.npy', allow_pickle=True)
# Function to get recommendations
def get_recommendations(img, n_neighbors=5):
img_path = "temp_input_image.png"
img.save(img_path)
img_features = extract_features(img_path, model)
distances, indices = nbrs.kneighbors([img_features])
recommended_images = [image_paths[idx] for idx in indices.flatten()]
recommended_model_names = [model_names[idx] for idx in indices.flatten()]
recommended_distances = distances.flatten()
return [(Image.open(img_path), f'{name}, Distance: {dist:.2f}') for img_path, name, dist in zip(recommended_images, recommended_model_names, recommended_distances)]
# Gradio interface
interface = gr.Interface(
fn=get_recommendations,
inputs=gr.Image(type="pil"),
outputs=gr.Gallery(label="Recommended Images"),
title="Image Recommendation System",
description="Upload an image and get similar images with their model names and distances."
)
if __name__ == "__main__":
interface.launch()