# -*- coding: utf-8 -*-
"""FinalProject.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1_wYfP0IRdb9fpc2zvbg8IqdXGx1dTo7X
"""

!pip install datasets transformers torch torchvision faiss-cpu gradio

from datasets import load_dataset
from PIL import Image

# Load dataset from Hugging Face
dataset = load_dataset("lirus18/deepfashion", split="train")

# Show one image
from IPython.display import display

image = dataset[0]['image']
display(image)

from transformers import CLIPProcessor, CLIPModel
import torch

# Load the CLIP model and processor
model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")

# Use GPU if available
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)

import numpy as np

image_vectors = []
image_indices = []

# Use a subset (you can increase to 1000+ later)
N = 500

for i in range(N):
    image = dataset[i]['image'].convert("RGB")
    inputs = processor(images=image, return_tensors="pt").to(device)
    with torch.no_grad():
        embedding = model.get_image_features(**inputs)
    image_vectors.append(embedding.cpu().numpy().squeeze())
    image_indices.append(i)  # Store index for later retrieval

image_vectors = np.array(image_vectors)

from sklearn.metrics.pairwise import cosine_similarity

def find_similar(user_image, top_k=3, exclude_index=None):
    # Embed the user image
    inputs = processor(images=user_image.convert("RGB"), return_tensors="pt").to(device)
    with torch.no_grad():
        query_vec = model.get_image_features(**inputs).cpu().numpy()

    # Compute cosine similarity
    sims = cosine_similarity(query_vec, image_vectors)[0]

    # Exclude the query image itself
    if exclude_index is not None:
        sims[exclude_index] = -1  # Force low similarity

    # Get top K similar image indices
    top_idx = sims.argsort()[-top_k:][::-1]

    return [dataset[image_indices[i]]['image'] for i in top_idx]

from IPython.display import display

query_index = 10
query_image = dataset[query_index]['image']
display(query_image)

similar_images = find_similar(query_image, exclude_index=query_index)

for img in similar_images:
    display(img)

from diffusers import StableDiffusionImg2ImgPipeline
import torch

# Load Stable Diffusion (this might take a while)
pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
).to(device)

def generate_outfit_from_image(input_image):
    prompt = "fashion outfit design inspired by the clothing item"
    init_image = input_image.resize((512, 512))
    generated = pipe(prompt=prompt, image=init_image, strength=0.7, guidance_scale=7.5)
    return generated.images[0]

from PIL import ImageChops

def recommend_from_upload(uploaded_image):
    # Step 1: Compare uploaded image to all dataset images and find the most identical one
    uploaded_image = uploaded_image.convert("RGB")
    closest_idx = None
    for i in range(len(image_indices)):
        dataset_image = dataset[image_indices[i]]['image'].convert("RGB")
        if ImageChops.difference(dataset_image, uploaded_image).getbbox() is None:
            closest_idx = i
            break

    # Step 2: Get top 3 similar, excluding the identical one if found
    similar_imgs = find_similar(uploaded_image, top_k=3, exclude_index=closest_idx)

    # Step 3: Generate 1 new outfit (placeholder for now)
    generated_img = generate_outfit_from_image(uploaded_image)

    return [uploaded_image] + similar_imgs + [generated_img]

import gradio as gr
import os

# Prepare example paths
example_images = [
    ["/content/fashion_examples/new1.jpg"],
    ["/content/fashion_examples/newnew.jpg"],
    ["/content/fashion_examples/newoutfit.jpg"],
]

# Build the interface
demo = gr.Interface(
    fn=recommend_from_upload,
    inputs=gr.Image(type="pil", label="Upload a clothing item"),
    outputs=[
        gr.Image(label="Your Input"),
        gr.Image(label="Similar Item 1"),
        gr.Image(label="Similar Item 2"),
        gr.Image(label="Similar Item 3"),
        gr.Image(label="Generated New Outfit"),
    ],
    title="👗 Fashion Outfit Recommender",
    description="Upload your own image *or* click an example to get 3 similar items + 1 AI-generated outfit.",
    examples=example_images
)

demo.launch()