Spaces:
Running
on
Zero
Running
on
Zero
File size: 1,464 Bytes
4ebb492 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 |
import torch
import gradio as gr
from PIL import Image
from transformers import AutoProcessor, SiglipModel
import faiss
import numpy as np
from huggingface_hub import hf_hub_download
from datasets import load_dataset
hf_hub_download("merve/siglip-faiss-wikiart", "siglip_new.index", local_dir="./")
index = faiss.read_index("./siglip_new.index")
dataset = load_dataset("huggan/wikiart")
device = torch.device('cuda' if torch.cuda.is_available() else "cpu")
dataset = dataset.with_format("torch", device=device)
processor = AutoProcessor.from_pretrained("nielsr/siglip-base-patch16-224")
model = SiglipModel.from_pretrained("nielsr/siglip-base-patch16-224").to(device)
def extract_features_siglip(image):
with torch.no_grad():
inputs = processor(images=image, return_tensors="pt").to(device)
image_features = model.get_image_features(**inputs)
return image_features
def infer(input_image):
input_features = extract_features_siglip(input_image)
input_features = input_features.detach().cpu().numpy()
input_features = np.float32(input_features)
faiss.normalize_L2(input_features)
distances, indices = index2.search(input_features, 9)
gallery_output = []
for i,v in enumerate(indices[0]):
sim = -distances[0][i]
img_resized = dataset["train"][int(v)]['image']
gallery_output.append(img_resized)
return gallery_output
gr.Interface(infer, "sketchpad", "gallery", title="Draw to Search Art 🖼️").launch()
|