Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -3,30 +3,19 @@ import streamlit as st
|
|
3 |
from PIL import Image
|
4 |
import torch
|
5 |
from torchvision import transforms, models
|
6 |
-
import numpy as np
|
7 |
-
from groq import Groq
|
8 |
|
9 |
-
# Set up environment
|
10 |
os.environ["GROQ_API_KEY"] = "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"
|
11 |
|
12 |
-
#
|
13 |
-
client = Groq(api_key=os.environ.get("GROQ_API_KEY"))
|
14 |
-
|
15 |
-
# Load Pretrained Models
|
16 |
@st.cache_resource
|
17 |
-
def
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
# Pretrained DenseNet (CheXNet) for normal/abnormal classification
|
23 |
-
chexnet_model = models.densenet121(pretrained=True)
|
24 |
-
chexnet_model.classifier = torch.nn.Linear(1024, 2) # Normal, Abnormal
|
25 |
-
chexnet_model.eval()
|
26 |
-
|
27 |
-
return organ_model, chexnet_model
|
28 |
|
29 |
-
organ_model
|
30 |
|
31 |
# Image Preprocessing
|
32 |
def preprocess_image(image):
|
@@ -37,35 +26,17 @@ def preprocess_image(image):
|
|
37 |
])
|
38 |
return transform(image).unsqueeze(0)
|
39 |
|
40 |
-
#
|
41 |
-
def get_ai_insights(text_prompt):
|
42 |
-
try:
|
43 |
-
response = client.chat.completions.create(
|
44 |
-
messages=[{"role": "user", "content": text_prompt}],
|
45 |
-
model="llama-3.3-70b-versatile"
|
46 |
-
)
|
47 |
-
return response.choices[0].message.content
|
48 |
-
except Exception as e:
|
49 |
-
return f"Error: {e}"
|
50 |
-
|
51 |
-
# Predict Organ
|
52 |
def predict_organ(image):
|
53 |
with torch.no_grad():
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
return prediction
|
58 |
-
|
59 |
-
# Predict Normal/Abnormal
|
60 |
-
def predict_normal_abnormal(image):
|
61 |
-
with torch.no_grad():
|
62 |
-
output = chexnet_model(preprocess_image(image))
|
63 |
-
classes = ["Normal", "Abnormal"]
|
64 |
prediction = classes[output.argmax().item()]
|
65 |
return prediction
|
66 |
|
67 |
# Streamlit App
|
68 |
-
st.title("
|
69 |
st.sidebar.title("Navigation")
|
70 |
task = st.sidebar.radio("Select a task", ["Upload X-ray", "AI Insights"])
|
71 |
|
@@ -77,24 +48,12 @@ if task == "Upload X-ray":
|
|
77 |
st.image(image, caption="Uploaded X-ray", use_column_width=True)
|
78 |
|
79 |
# Predict Organ
|
80 |
-
st.subheader("Step 1: Identify the Organ")
|
81 |
organ = predict_organ(image)
|
82 |
st.write(f"Predicted Organ: **{organ}**")
|
83 |
|
84 |
-
# Predict Normal/Abnormal
|
85 |
-
st.subheader("Step 2: Analyze the X-ray")
|
86 |
-
classification = predict_normal_abnormal(image)
|
87 |
-
st.write(f"X-ray Status: **{classification}**")
|
88 |
-
|
89 |
-
if classification == "Abnormal":
|
90 |
-
st.subheader("Step 3: AI-Based Insights")
|
91 |
-
ai_prompt = f"Explain why this X-ray of the {organ} is abnormal."
|
92 |
-
insights = get_ai_insights(ai_prompt)
|
93 |
-
st.write(insights)
|
94 |
-
|
95 |
elif task == "AI Insights":
|
96 |
st.subheader("Ask AI")
|
97 |
user_input = st.text_area("Enter your query for AI insights")
|
98 |
if user_input:
|
99 |
-
|
100 |
-
st.write(response)
|
|
|
3 |
from PIL import Image
|
4 |
import torch
|
5 |
from torchvision import transforms, models
|
|
|
|
|
6 |
|
7 |
+
# Set up environment variable for Groq API
|
8 |
os.environ["GROQ_API_KEY"] = "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"
|
9 |
|
10 |
+
# Load Pretrained Model for Organ Recognition
|
|
|
|
|
|
|
11 |
@st.cache_resource
|
12 |
+
def load_organ_model():
|
13 |
+
model = models.resnet18(pretrained=True) # ResNet18 pretrained model
|
14 |
+
model.fc = torch.nn.Linear(model.fc.in_features, 4) # Modify for 4 classes
|
15 |
+
model.eval()
|
16 |
+
return model
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
|
18 |
+
organ_model = load_organ_model()
|
19 |
|
20 |
# Image Preprocessing
|
21 |
def preprocess_image(image):
|
|
|
26 |
])
|
27 |
return transform(image).unsqueeze(0)
|
28 |
|
29 |
+
# Organ Recognition Prediction
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
def predict_organ(image):
|
31 |
with torch.no_grad():
|
32 |
+
input_tensor = preprocess_image(image)
|
33 |
+
output = organ_model(input_tensor)
|
34 |
+
classes = ["Lungs", "Heart", "Spine", "Other"] # Example organ classes
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
prediction = classes[output.argmax().item()]
|
36 |
return prediction
|
37 |
|
38 |
# Streamlit App
|
39 |
+
st.title("X-ray Organ Recognition App")
|
40 |
st.sidebar.title("Navigation")
|
41 |
task = st.sidebar.radio("Select a task", ["Upload X-ray", "AI Insights"])
|
42 |
|
|
|
48 |
st.image(image, caption="Uploaded X-ray", use_column_width=True)
|
49 |
|
50 |
# Predict Organ
|
51 |
+
st.subheader("Step 1: Identify the Organ in the X-ray")
|
52 |
organ = predict_organ(image)
|
53 |
st.write(f"Predicted Organ: **{organ}**")
|
54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
elif task == "AI Insights":
|
56 |
st.subheader("Ask AI")
|
57 |
user_input = st.text_area("Enter your query for AI insights")
|
58 |
if user_input:
|
59 |
+
st.write("AI insights will be generated here.") # Placeholder for AI logic
|
|