Spaces:
Sleeping
Sleeping
Update src/streamlit_app.py
Browse files- src/streamlit_app.py +92 -93
src/streamlit_app.py
CHANGED
@@ -7,30 +7,12 @@ import numpy as np
|
|
7 |
import chromadb
|
8 |
import requests
|
9 |
import tempfile
|
10 |
-
from tqdm import tqdm
|
11 |
-
|
12 |
-
# Get a temporary directory (automatically cleaned up after runtime ends)
|
13 |
-
temp_dir = tempfile.gettempdir()
|
14 |
-
demo_dir = os.path.join(temp_dir, "demo_images")
|
15 |
-
os.makedirs(demo_dir, exist_ok=True)
|
16 |
-
|
17 |
-
print(f"Saving images to: {demo_dir}")
|
18 |
-
|
19 |
-
# Download 50 high-resolution images (1024x768)
|
20 |
-
for i in tqdm(range(50), desc="Downloading images"):
|
21 |
-
url = f"https://picsum.photos/seed/{i}/1024/768"
|
22 |
-
response = requests.get(url)
|
23 |
-
if response.status_code == 200:
|
24 |
-
with open(os.path.join(demo_dir, f"img_{i+1:02}.jpg"), "wb") as f:
|
25 |
-
f.write(response.content)
|
26 |
-
else:
|
27 |
-
print(f"Failed to download image {i+1}")
|
28 |
-
|
29 |
-
|
30 |
|
31 |
# ----- Setup -----
|
32 |
CACHE_DIR = tempfile.gettempdir()
|
33 |
CHROMA_PATH = os.path.join(CACHE_DIR, "chroma_db")
|
|
|
|
|
34 |
|
35 |
# ----- Load CLIP Model -----
|
36 |
if 'model' not in st.session_state:
|
@@ -39,9 +21,6 @@ if 'model' not in st.session_state:
|
|
39 |
st.session_state.model = model
|
40 |
st.session_state.preprocess = preprocess
|
41 |
st.session_state.device = device
|
42 |
-
st.session_state.demo_images = []
|
43 |
-
st.session_state.demo_image_paths = []
|
44 |
-
st.session_state.user_images = []
|
45 |
|
46 |
# ----- Initialize ChromaDB -----
|
47 |
if 'chroma_client' not in st.session_state:
|
@@ -53,19 +32,36 @@ if 'chroma_client' not in st.session_state:
|
|
53 |
name="user_images", metadata={"hnsw:space": "cosine"}
|
54 |
)
|
55 |
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
|
67 |
embeddings, ids, metadatas = [], [], []
|
68 |
-
for i, img in enumerate(
|
69 |
img_tensor = st.session_state.preprocess(img).unsqueeze(0).to(st.session_state.device)
|
70 |
with torch.no_grad():
|
71 |
embedding = st.session_state.model.encode_image(img_tensor).cpu().numpy().flatten()
|
@@ -74,78 +70,81 @@ if not st.session_state.get("demo_images_loaded", False):
|
|
74 |
metadatas.append({"path": demo_image_paths[i]})
|
75 |
|
76 |
st.session_state.demo_collection.add(embeddings=embeddings, ids=ids, metadatas=metadatas)
|
77 |
-
st.session_state.
|
|
|
|
|
78 |
|
79 |
-
|
80 |
-
st.title("π CLIP Image Search (Text & Image)")
|
81 |
-
mode = st.radio("Choose dataset to search in:", ("Demo Images", "My Uploaded Images"))
|
82 |
-
query_type = st.radio("Query type:", ("Image", "Text"))
|
83 |
|
84 |
-
# -----
|
85 |
-
if
|
86 |
-
uploaded = st.file_uploader("Upload your images", type=[
|
87 |
if uploaded:
|
88 |
-
st.session_state.user_images = []
|
89 |
st.session_state.user_collection.delete(ids=[
|
90 |
str(i) for i in range(st.session_state.user_collection.count())
|
91 |
])
|
92 |
-
|
93 |
for i, file in enumerate(uploaded):
|
94 |
img = Image.open(file).convert("RGB")
|
95 |
-
|
96 |
|
97 |
img_tensor = st.session_state.preprocess(img).unsqueeze(0).to(st.session_state.device)
|
98 |
with torch.no_grad():
|
99 |
embedding = st.session_state.model.encode_image(img_tensor).cpu().numpy().flatten()
|
100 |
-
|
101 |
st.session_state.user_collection.add(
|
102 |
-
embeddings=[embedding],
|
103 |
-
ids=[str(i)],
|
104 |
-
metadatas=[{"index": i}]
|
105 |
)
|
106 |
|
107 |
-
st.
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
query_embedding = st.session_state.model.encode_text(tokens).cpu().numpy().flatten()
|
125 |
-
|
126 |
-
# ----- Run Search -----
|
127 |
-
if query_embedding is not None:
|
128 |
-
if mode == "Demo Images":
|
129 |
-
collection = st.session_state.demo_collection
|
130 |
-
images = st.session_state.demo_images
|
131 |
-
else:
|
132 |
-
collection = st.session_state.user_collection
|
133 |
-
images = st.session_state.user_images
|
134 |
-
|
135 |
-
if collection.count() > 0:
|
136 |
-
results = collection.query(
|
137 |
-
query_embeddings=[query_embedding],
|
138 |
-
n_results=min(5, collection.count())
|
139 |
-
)
|
140 |
-
ids = results["ids"][0]
|
141 |
-
distances = results["distances"][0]
|
142 |
-
similarities = [1 - d for d in distances]
|
143 |
-
|
144 |
-
st.subheader("Top Matches")
|
145 |
-
cols = st.columns(5)
|
146 |
-
for i, (img_id, sim) in enumerate(zip(ids, similarities)):
|
147 |
-
with cols[i]:
|
148 |
-
idx = int(img_id)
|
149 |
-
st.image(images[idx], caption=f"Sim: {sim:.3f}", width=150)
|
150 |
else:
|
151 |
-
st.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
import chromadb
|
8 |
import requests
|
9 |
import tempfile
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
# ----- Setup -----
|
12 |
CACHE_DIR = tempfile.gettempdir()
|
13 |
CHROMA_PATH = os.path.join(CACHE_DIR, "chroma_db")
|
14 |
+
DEMO_DIR = os.path.join(CACHE_DIR, "demo_images")
|
15 |
+
os.makedirs(DEMO_DIR, exist_ok=True)
|
16 |
|
17 |
# ----- Load CLIP Model -----
|
18 |
if 'model' not in st.session_state:
|
|
|
21 |
st.session_state.model = model
|
22 |
st.session_state.preprocess = preprocess
|
23 |
st.session_state.device = device
|
|
|
|
|
|
|
24 |
|
25 |
# ----- Initialize ChromaDB -----
|
26 |
if 'chroma_client' not in st.session_state:
|
|
|
32 |
name="user_images", metadata={"hnsw:space": "cosine"}
|
33 |
)
|
34 |
|
35 |
+
st.title("π CLIP-Based Image Search")
|
36 |
+
|
37 |
+
# Dataset selection
|
38 |
+
col1, col2 = st.columns(2)
|
39 |
+
use_demo = col1.button("π¦ Use Demo Images")
|
40 |
+
upload_own = col2.button("π€ Upload Your Images")
|
41 |
+
|
42 |
+
dataset_loaded = False
|
43 |
+
dataset_name = None
|
44 |
+
|
45 |
+
# ----- Handle Demo Images -----
|
46 |
+
if use_demo:
|
47 |
+
with st.spinner("Downloading and indexing demo images..."):
|
48 |
+
st.session_state.demo_collection.delete(ids=[str(i) for i in range(50)])
|
49 |
+
demo_image_paths = []
|
50 |
+
demo_images = []
|
51 |
+
|
52 |
+
for i in range(50):
|
53 |
+
path = os.path.join(DEMO_DIR, f"img_{i+1:02}.jpg")
|
54 |
+
if not os.path.exists(path):
|
55 |
+
url = f"https://picsum.photos/seed/{i}/1024/768"
|
56 |
+
response = requests.get(url)
|
57 |
+
if response.status_code == 200:
|
58 |
+
with open(path, "wb") as f:
|
59 |
+
f.write(response.content)
|
60 |
+
demo_image_paths.append(path)
|
61 |
+
demo_images.append(Image.open(path).convert("RGB"))
|
62 |
|
63 |
embeddings, ids, metadatas = [], [], []
|
64 |
+
for i, img in enumerate(demo_images):
|
65 |
img_tensor = st.session_state.preprocess(img).unsqueeze(0).to(st.session_state.device)
|
66 |
with torch.no_grad():
|
67 |
embedding = st.session_state.model.encode_image(img_tensor).cpu().numpy().flatten()
|
|
|
70 |
metadatas.append({"path": demo_image_paths[i]})
|
71 |
|
72 |
st.session_state.demo_collection.add(embeddings=embeddings, ids=ids, metadatas=metadatas)
|
73 |
+
st.session_state.demo_images = demo_images
|
74 |
+
dataset_loaded = True
|
75 |
+
dataset_name = "demo"
|
76 |
|
77 |
+
st.success("Demo images loaded!")
|
|
|
|
|
|
|
78 |
|
79 |
+
# ----- Handle User Uploads -----
|
80 |
+
if upload_own:
|
81 |
+
uploaded = st.file_uploader("Upload your images", type=["jpg", "jpeg", "png"], accept_multiple_files=True)
|
82 |
if uploaded:
|
|
|
83 |
st.session_state.user_collection.delete(ids=[
|
84 |
str(i) for i in range(st.session_state.user_collection.count())
|
85 |
])
|
86 |
+
user_images = []
|
87 |
for i, file in enumerate(uploaded):
|
88 |
img = Image.open(file).convert("RGB")
|
89 |
+
user_images.append(img)
|
90 |
|
91 |
img_tensor = st.session_state.preprocess(img).unsqueeze(0).to(st.session_state.device)
|
92 |
with torch.no_grad():
|
93 |
embedding = st.session_state.model.encode_image(img_tensor).cpu().numpy().flatten()
|
|
|
94 |
st.session_state.user_collection.add(
|
95 |
+
embeddings=[embedding], ids=[str(i)], metadatas=[{"index": i}]
|
|
|
|
|
96 |
)
|
97 |
|
98 |
+
st.session_state.user_images = user_images
|
99 |
+
st.success(f"{len(user_images)} images uploaded.")
|
100 |
+
dataset_loaded = True
|
101 |
+
dataset_name = "user"
|
102 |
+
|
103 |
+
# ----- Search UI -----
|
104 |
+
if dataset_loaded:
|
105 |
+
st.subheader("Search Section")
|
106 |
+
query_type = st.radio("Search by:", ("Text", "Image"))
|
107 |
+
|
108 |
+
query_embedding = None
|
109 |
+
if query_type == "Text":
|
110 |
+
text_query = st.text_input("Enter search text:")
|
111 |
+
if text_query:
|
112 |
+
tokens = clip.tokenize([text_query]).to(st.session_state.device)
|
113 |
+
with torch.no_grad():
|
114 |
+
query_embedding = st.session_state.model.encode_text(tokens).cpu().numpy().flatten()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
115 |
else:
|
116 |
+
img_file = st.file_uploader("Upload query image", type=["jpg", "jpeg", "png"])
|
117 |
+
if img_file:
|
118 |
+
query_img = Image.open(img_file).convert("RGB")
|
119 |
+
st.image(query_img, caption="Query Image", width=200)
|
120 |
+
img_tensor = st.session_state.preprocess(query_img).unsqueeze(0).to(st.session_state.device)
|
121 |
+
with torch.no_grad():
|
122 |
+
query_embedding = st.session_state.model.encode_image(img_tensor).cpu().numpy().flatten()
|
123 |
+
|
124 |
+
# ----- Perform Search -----
|
125 |
+
if query_embedding is not None:
|
126 |
+
if dataset_name == "demo":
|
127 |
+
collection = st.session_state.demo_collection
|
128 |
+
images = st.session_state.demo_images
|
129 |
+
else:
|
130 |
+
collection = st.session_state.user_collection
|
131 |
+
images = st.session_state.user_images
|
132 |
+
|
133 |
+
if collection.count() > 0:
|
134 |
+
results = collection.query(
|
135 |
+
query_embeddings=[query_embedding],
|
136 |
+
n_results=min(5, collection.count())
|
137 |
+
)
|
138 |
+
ids = results["ids"][0]
|
139 |
+
distances = results["distances"][0]
|
140 |
+
similarities = [1 - d for d in distances]
|
141 |
+
|
142 |
+
st.subheader("Top Matches")
|
143 |
+
cols = st.columns(len(ids))
|
144 |
+
for i, (img_id, sim) in enumerate(zip(ids, similarities)):
|
145 |
+
with cols[i]:
|
146 |
+
st.image(images[int(img_id)], caption=f"Sim: {sim:.3f}", width=150)
|
147 |
+
else:
|
148 |
+
st.warning("No images in the collection.")
|
149 |
+
else:
|
150 |
+
st.info("Please click on one of the options above to load a dataset.")
|