main.py

import numpy as np
import faiss
import torch
from torchvision.transforms import (
    Compose,
    Resize,
    ToTensor,
    Normalize,
    InterpolationMode,
)
from PIL import Image
import gradio as gr

(ys,) = np.load("embs.npz").values()
model = torch.load(
    "style-extractor-v0.2.0.ckpt",
    map_location="cpu",
)
with open("urls.txt") as f:
    urls = f.read().splitlines()
assert len(urls) == len(ys)
d = ys.shape[1]
index = faiss.IndexFlatL2(d)
index.is_trained
index.add(ys)
tf = Compose(
    [
        Resize(
            size=336,
            interpolation=InterpolationMode.BICUBIC,
            max_size=None,
            antialias=True,
        ),
        ToTensor(),
        Normalize(mean=[0.4850, 0.4560, 0.4060], std=[0.2290, 0.2240, 0.2250]),
    ]
)


def get_emb(im: Image):
    model.eval()
    with torch.no_grad():
        return model(tf(im).unsqueeze(0))


n_outputs = 50
row_size = 5


def f(im):
    D, I = index.search(get_emb(im), n_outputs)
    return [f"Distance: {d}\n![]({urls[i]})" for d, i in zip(D[0], I[0])]


with gr.Blocks() as demo:
    gr.Markdown(
        "# Style Similarity Search\n\nFind artworks with a similar style from a small database (10k artists * 6img/artist)"
    )
    img = gr.Image(type="pil", label="Query", height=500)
    btn = gr.Button(variant="primary", value="search")
    outputs = []
    for i in range(-(n_outputs // (-row_size))):
        with gr.Row():
            for _ in range(min(row_size, n_outputs - i * row_size)):
                outputs.append(gr.Markdown(label=f"#{len(outputs) + 1}"))
    btn.click(f, img, outputs)
demo.launch()