Spaces:
Runtime error
Runtime error
Commit
·
d6c88ae
1
Parent(s):
b68c187
Update app.py
Browse files
app.py
CHANGED
@@ -1,20 +1,71 @@
|
|
|
|
1 |
import streamlit as st
|
2 |
-
|
|
|
|
|
|
|
|
|
3 |
from PIL import Image
|
|
|
|
|
4 |
|
5 |
-
|
|
|
6 |
|
7 |
-
|
|
|
|
|
|
|
8 |
|
9 |
-
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
-
|
12 |
-
|
|
|
|
|
|
|
13 |
|
14 |
-
|
15 |
-
|
16 |
-
|
|
|
|
|
|
|
|
|
17 |
|
18 |
-
|
19 |
-
|
20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from datasets import load_dataset
|
2 |
import streamlit as st
|
3 |
+
import torch
|
4 |
+
from transformers import AutoTokenizer, AutoModel
|
5 |
+
import faiss
|
6 |
+
import numpy as np
|
7 |
+
import wget
|
8 |
from PIL import Image
|
9 |
+
from io import BytesIO
|
10 |
+
from sentence_transformers import SentenceTransformer
|
11 |
|
12 |
+
# dataset = load_dataset("nlphuji/flickr30k", streaming=True)
|
13 |
+
# df = pd.DataFrame.from_dict(dataset["train"])
|
14 |
|
15 |
+
# Load the pre-trained sentence encoder
|
16 |
+
model_name = "sentence-transformers/paraphrase-multilingual-mpnet-base-v2"
|
17 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
18 |
+
model = SentenceTransformer(model_name)
|
19 |
|
20 |
+
# # Load the pre-trained image model
|
21 |
+
# image_model_name = 'image_model.ckpt'
|
22 |
+
# image_model_url = 'https://huggingface.co/models/flax-community/deit-tiny-random/images/vqvae.png'
|
23 |
+
# wget.download(image_model_url, image_model_name)
|
24 |
+
# image_model = torch.load(image_model_name, map_location=torch.device('cpu'))
|
25 |
+
# image_model.eval()
|
26 |
|
27 |
+
# Load the FAISS index
|
28 |
+
index_name = 'index.faiss'
|
29 |
+
index_url = 'https://huggingface.co/models/flax-community/deit-tiny-random/faiss_files/faiss.index'
|
30 |
+
wget.download(index_url, index_name)
|
31 |
+
index = faiss.read_index(index_name)
|
32 |
|
33 |
+
# Map the image ids to the corresponding image URLs
|
34 |
+
image_map_name = 'image_map.json'
|
35 |
+
image_map_url = 'https://huggingface.co/models/flax-community/deit-tiny-random/faiss_files/image_map.json'
|
36 |
+
wget.download(image_map_url, image_map_name)
|
37 |
+
image_map = {}
|
38 |
+
with open(image_map_name, 'r') as f:
|
39 |
+
image_map = json.load(f)
|
40 |
|
41 |
+
def search(query, k=5):
|
42 |
+
# Encode the query
|
43 |
+
query_tokens = tokenizer.encode(query, return_tensors='pt')
|
44 |
+
query_embedding = model.encode(query_tokens).detach().numpy()
|
45 |
+
|
46 |
+
# Search for the nearest neighbors in the FAISS index
|
47 |
+
D, I = index.search(query_embedding, k)
|
48 |
+
|
49 |
+
# Map the image ids to the corresponding image URLs
|
50 |
+
image_urls = []
|
51 |
+
for i in I[0]:
|
52 |
+
image_id = str(i)
|
53 |
+
image_url = image_map[image_id]
|
54 |
+
image_urls.append(image_url)
|
55 |
+
|
56 |
+
return image_urls
|
57 |
+
|
58 |
+
st.title("Image Search App")
|
59 |
+
|
60 |
+
query = st.text_input("Enter your search query here:")
|
61 |
+
if st.button("Search"):
|
62 |
+
if query:
|
63 |
+
image_urls = search(query)
|
64 |
+
|
65 |
+
# Display the images
|
66 |
+
st.image(image_urls, width=200)
|
67 |
+
|
68 |
+
if __name__ == '__main__':
|
69 |
+
st.set_page_config(page_title='Image Search App', layout='wide')
|
70 |
+
st.cache(allow_output_mutation=True)
|
71 |
+
run_app()
|