Update app.py
Browse files
app.py
CHANGED
@@ -11,6 +11,8 @@ import cv2
|
|
11 |
from inference_sdk import InferenceHTTPClient
|
12 |
import matplotlib.pyplot as plt
|
13 |
import base64
|
|
|
|
|
14 |
|
15 |
# Load model and tokenizer
|
16 |
@st.cache_resource
|
@@ -30,88 +32,77 @@ def load_data():
|
|
30 |
|
31 |
data = load_data()
|
32 |
|
33 |
-
|
34 |
-
|
|
|
|
|
|
|
|
|
35 |
def download_and_process_image(image_url):
|
36 |
try:
|
37 |
response = requests.get(image_url)
|
38 |
-
response.raise_for_status()
|
39 |
image = Image.open(BytesIO(response.content))
|
40 |
-
|
41 |
-
# Convert image to RGB mode if it's in RGBA mode
|
42 |
if image.mode == 'RGBA':
|
43 |
image = image.convert('RGB')
|
44 |
-
|
45 |
return image
|
46 |
-
except requests.RequestException as e:
|
47 |
-
st.error(f"Error downloading image: {e}")
|
48 |
-
return None
|
49 |
except Exception as e:
|
50 |
-
st.error(f"Error processing image: {e}")
|
51 |
return None
|
52 |
|
53 |
-
def
|
54 |
-
image_tensor = preprocess_val(image).unsqueeze(0).to(device)
|
55 |
-
with torch.no_grad():
|
56 |
-
image_features = model.encode_image(image_tensor)
|
57 |
-
image_features /= image_features.norm(dim=-1, keepdim=True)
|
58 |
-
return image_features.cpu().numpy()
|
59 |
-
|
60 |
-
def setup_roboflow_client(api_key):
|
61 |
-
return InferenceHTTPClient(
|
62 |
-
api_url="https://outline.roboflow.com",
|
63 |
-
api_key=api_key
|
64 |
-
)
|
65 |
-
|
66 |
-
def segment_image(image_path, client):
|
67 |
try:
|
68 |
-
# 이미지 파일 읽기
|
69 |
with open(image_path, "rb") as image_file:
|
70 |
image_data = image_file.read()
|
71 |
|
72 |
-
# 이미지를 base64로 인코딩
|
73 |
encoded_image = base64.b64encode(image_data).decode('utf-8')
|
74 |
|
75 |
-
# 원본 이미지 로드
|
76 |
image = cv2.imread(image_path)
|
77 |
image = cv2.resize(image, (800, 600))
|
78 |
mask = np.zeros(image.shape, dtype=np.uint8)
|
79 |
|
80 |
-
# Roboflow API 호출
|
81 |
results = client.infer(encoded_image, model_id="closet/1")
|
82 |
|
83 |
-
# 결과가 이미 딕셔너리인 경우 JSON 파싱 단계 제거
|
84 |
if isinstance(results, dict):
|
85 |
predictions = results.get('predictions', [])
|
86 |
else:
|
87 |
-
# 문자열인 경우에만 JSON 파싱
|
88 |
predictions = json.loads(results).get('predictions', [])
|
89 |
|
|
|
90 |
if predictions:
|
91 |
for prediction in predictions:
|
92 |
points = prediction['points']
|
93 |
pts = np.array([[p['x'], p['y']] for p in points], np.int32)
|
94 |
-
scale_x = image.shape[1] / results
|
95 |
-
scale_y = image.shape[0] / results
|
96 |
pts = pts * [scale_x, scale_y]
|
97 |
pts = pts.astype(np.int32)
|
98 |
pts = pts.reshape((-1, 1, 2))
|
99 |
-
cv2.fillPoly(mask, [pts], color=(255, 255, 255))
|
|
|
|
|
|
|
|
|
100 |
|
101 |
segmented_image = cv2.bitwise_and(image, mask)
|
102 |
else:
|
103 |
st.warning("No predictions found in the image. Returning original image.")
|
104 |
segmented_image = image
|
105 |
|
106 |
-
return Image.fromarray(cv2.cvtColor(segmented_image, cv2.COLOR_BGR2RGB))
|
107 |
except Exception as e:
|
108 |
st.error(f"Error in segmentation: {str(e)}")
|
109 |
-
|
110 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
111 |
|
112 |
@st.cache_data
|
113 |
def process_database_cached(data):
|
114 |
-
database_embeddings = []
|
115 |
database_info = []
|
116 |
for item in data:
|
117 |
image_url = item['이미지 링크'][0]
|
@@ -121,7 +112,6 @@ def process_database_cached(data):
|
|
121 |
if image is None:
|
122 |
continue
|
123 |
|
124 |
-
# Save the image temporarily
|
125 |
temp_path = f"temp_{product_id}.jpg"
|
126 |
image.save(temp_path, 'JPEG')
|
127 |
|
@@ -140,17 +130,42 @@ def process_database_cached(data):
|
|
140 |
|
141 |
def process_database(client, data):
|
142 |
database_info = process_database_cached(data)
|
143 |
-
|
|
|
144 |
|
|
|
145 |
for item in database_info:
|
146 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
147 |
embedding = get_image_embedding(segmented_image)
|
148 |
database_embeddings.append(embedding)
|
|
|
149 |
|
150 |
return np.vstack(database_embeddings), database_info
|
151 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
152 |
# Streamlit app
|
153 |
-
st.title("Fashion Search App with Segmentation")
|
154 |
|
155 |
# API Key input
|
156 |
api_key = st.text_input("Enter your Roboflow API Key", type="password")
|
@@ -168,17 +183,18 @@ if api_key:
|
|
168 |
|
169 |
if st.button('Find Similar Items'):
|
170 |
with st.spinner('Processing...'):
|
171 |
-
# Save uploaded image temporarily
|
172 |
temp_path = "temp_upload.jpg"
|
173 |
image.save(temp_path)
|
174 |
|
175 |
-
|
176 |
-
segmented_image = segment_image(temp_path, CLIENT)
|
177 |
st.image(segmented_image, caption='Segmented Image', use_column_width=True)
|
178 |
|
179 |
-
|
|
|
|
|
|
|
180 |
query_embedding = get_image_embedding(segmented_image)
|
181 |
-
similar_images = find_similar_images(query_embedding)
|
182 |
|
183 |
st.subheader("Similar Items:")
|
184 |
for img in similar_images:
|
@@ -192,5 +208,9 @@ if api_key:
|
|
192 |
st.write(f"Price: {img['info']['price']}")
|
193 |
st.write(f"Discount: {img['info']['discount']}%")
|
194 |
st.write(f"Similarity: {img['similarity']:.2f}")
|
|
|
|
|
|
|
|
|
195 |
else:
|
196 |
st.warning("Please enter your Roboflow API Key to use the app.")
|
|
|
11 |
from inference_sdk import InferenceHTTPClient
|
12 |
import matplotlib.pyplot as plt
|
13 |
import base64
|
14 |
+
import os
|
15 |
+
import pickle
|
16 |
|
17 |
# Load model and tokenizer
|
18 |
@st.cache_resource
|
|
|
32 |
|
33 |
data = load_data()
|
34 |
|
35 |
+
def setup_roboflow_client(api_key):
|
36 |
+
return InferenceHTTPClient(
|
37 |
+
api_url="https://outline.roboflow.com",
|
38 |
+
api_key=api_key
|
39 |
+
)
|
40 |
+
|
41 |
def download_and_process_image(image_url):
|
42 |
try:
|
43 |
response = requests.get(image_url)
|
44 |
+
response.raise_for_status()
|
45 |
image = Image.open(BytesIO(response.content))
|
|
|
|
|
46 |
if image.mode == 'RGBA':
|
47 |
image = image.convert('RGB')
|
|
|
48 |
return image
|
|
|
|
|
|
|
49 |
except Exception as e:
|
50 |
+
st.error(f"Error downloading/processing image: {str(e)}")
|
51 |
return None
|
52 |
|
53 |
+
def segment_image_and_get_categories(image_path, client):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
try:
|
|
|
55 |
with open(image_path, "rb") as image_file:
|
56 |
image_data = image_file.read()
|
57 |
|
|
|
58 |
encoded_image = base64.b64encode(image_data).decode('utf-8')
|
59 |
|
|
|
60 |
image = cv2.imread(image_path)
|
61 |
image = cv2.resize(image, (800, 600))
|
62 |
mask = np.zeros(image.shape, dtype=np.uint8)
|
63 |
|
|
|
64 |
results = client.infer(encoded_image, model_id="closet/1")
|
65 |
|
|
|
66 |
if isinstance(results, dict):
|
67 |
predictions = results.get('predictions', [])
|
68 |
else:
|
|
|
69 |
predictions = json.loads(results).get('predictions', [])
|
70 |
|
71 |
+
categories = []
|
72 |
if predictions:
|
73 |
for prediction in predictions:
|
74 |
points = prediction['points']
|
75 |
pts = np.array([[p['x'], p['y']] for p in points], np.int32)
|
76 |
+
scale_x = image.shape[1] / results.get('image', {}).get('width', 1)
|
77 |
+
scale_y = image.shape[0] / results.get('image', {}).get('height', 1)
|
78 |
pts = pts * [scale_x, scale_y]
|
79 |
pts = pts.astype(np.int32)
|
80 |
pts = pts.reshape((-1, 1, 2))
|
81 |
+
cv2.fillPoly(mask, [pts], color=(255, 255, 255))
|
82 |
+
|
83 |
+
category = prediction.get('class', 'Unknown')
|
84 |
+
confidence = prediction.get('confidence', 0)
|
85 |
+
categories.append(f"{category} ({confidence:.2f})")
|
86 |
|
87 |
segmented_image = cv2.bitwise_and(image, mask)
|
88 |
else:
|
89 |
st.warning("No predictions found in the image. Returning original image.")
|
90 |
segmented_image = image
|
91 |
|
92 |
+
return Image.fromarray(cv2.cvtColor(segmented_image, cv2.COLOR_BGR2RGB)), categories
|
93 |
except Exception as e:
|
94 |
st.error(f"Error in segmentation: {str(e)}")
|
95 |
+
return Image.open(image_path), []
|
96 |
+
|
97 |
+
def get_image_embedding(image):
|
98 |
+
image_tensor = preprocess_val(image).unsqueeze(0).to(device)
|
99 |
+
with torch.no_grad():
|
100 |
+
image_features = model.encode_image(image_tensor)
|
101 |
+
image_features /= image_features.norm(dim=-1, keepdim=True)
|
102 |
+
return image_features.cpu().numpy()
|
103 |
|
104 |
@st.cache_data
|
105 |
def process_database_cached(data):
|
|
|
106 |
database_info = []
|
107 |
for item in data:
|
108 |
image_url = item['이미지 링크'][0]
|
|
|
112 |
if image is None:
|
113 |
continue
|
114 |
|
|
|
115 |
temp_path = f"temp_{product_id}.jpg"
|
116 |
image.save(temp_path, 'JPEG')
|
117 |
|
|
|
130 |
|
131 |
def process_database(client, data):
|
132 |
database_info = process_database_cached(data)
|
133 |
+
cache_dir = "segmentation_cache"
|
134 |
+
os.makedirs(cache_dir, exist_ok=True)
|
135 |
|
136 |
+
database_embeddings = []
|
137 |
for item in database_info:
|
138 |
+
cache_file = os.path.join(cache_dir, f"{item['id']}_segmented.pkl")
|
139 |
+
|
140 |
+
if os.path.exists(cache_file):
|
141 |
+
with open(cache_file, 'rb') as f:
|
142 |
+
segmented_image, categories = pickle.load(f)
|
143 |
+
else:
|
144 |
+
segmented_image, categories = segment_image_and_get_categories(item['temp_path'], client)
|
145 |
+
with open(cache_file, 'wb') as f:
|
146 |
+
pickle.dump((segmented_image, categories), f)
|
147 |
+
|
148 |
embedding = get_image_embedding(segmented_image)
|
149 |
database_embeddings.append(embedding)
|
150 |
+
item['categories'] = categories
|
151 |
|
152 |
return np.vstack(database_embeddings), database_info
|
153 |
|
154 |
+
def find_similar_images(query_embedding, database_embeddings, database_info, top_k=5):
|
155 |
+
similarities = np.dot(database_embeddings, query_embedding.T).squeeze()
|
156 |
+
top_indices = np.argsort(similarities)[::-1][:top_k]
|
157 |
+
|
158 |
+
results = []
|
159 |
+
for idx in top_indices:
|
160 |
+
results.append({
|
161 |
+
'info': database_info[idx],
|
162 |
+
'similarity': similarities[idx]
|
163 |
+
})
|
164 |
+
|
165 |
+
return results
|
166 |
+
|
167 |
# Streamlit app
|
168 |
+
st.title("Fashion Search App with Segmentation and Category Detection")
|
169 |
|
170 |
# API Key input
|
171 |
api_key = st.text_input("Enter your Roboflow API Key", type="password")
|
|
|
183 |
|
184 |
if st.button('Find Similar Items'):
|
185 |
with st.spinner('Processing...'):
|
|
|
186 |
temp_path = "temp_upload.jpg"
|
187 |
image.save(temp_path)
|
188 |
|
189 |
+
segmented_image, input_categories = segment_image_and_get_categories(temp_path, CLIENT)
|
|
|
190 |
st.image(segmented_image, caption='Segmented Image', use_column_width=True)
|
191 |
|
192 |
+
st.subheader("Detected Categories in Input Image:")
|
193 |
+
for category in input_categories:
|
194 |
+
st.write(category)
|
195 |
+
|
196 |
query_embedding = get_image_embedding(segmented_image)
|
197 |
+
similar_images = find_similar_images(query_embedding, database_embeddings, database_info)
|
198 |
|
199 |
st.subheader("Similar Items:")
|
200 |
for img in similar_images:
|
|
|
208 |
st.write(f"Price: {img['info']['price']}")
|
209 |
st.write(f"Discount: {img['info']['discount']}%")
|
210 |
st.write(f"Similarity: {img['similarity']:.2f}")
|
211 |
+
|
212 |
+
st.write("Detected Categories:")
|
213 |
+
for category in img['info']['categories']:
|
214 |
+
st.write(category)
|
215 |
else:
|
216 |
st.warning("Please enter your Roboflow API Key to use the app.")
|