Update app.py
Browse files
app.py
CHANGED
@@ -24,8 +24,6 @@ from sklearn.cluster import DBSCAN
|
|
24 |
# Streamlit
|
25 |
import streamlit as st
|
26 |
st.set_option('deprecation.showPyplotGlobalUse', False)
|
27 |
-
st.set_page_config(page_title="Cavity Detection Tool", layout="wide")
|
28 |
-
# st.title("Cavity Detection Tool")
|
29 |
|
30 |
# Define function to plot the uploaded image
|
31 |
def plot_image(image, scale):
|
@@ -58,7 +56,7 @@ def plot_decomposed(decomposed):
|
|
58 |
plt.axis('off')
|
59 |
with colC: st.pyplot()
|
60 |
|
61 |
-
#
|
62 |
def cut(data0, wcs0, scale=1):
|
63 |
shape = data0.shape[0]
|
64 |
x0 = shape / 2
|
@@ -81,7 +79,7 @@ def cut(data0, wcs0, scale=1):
|
|
81 |
|
82 |
return data, wcs
|
83 |
|
84 |
-
|
85 |
@st.cache
|
86 |
def cut_n_predict(data, wcs, scale):
|
87 |
data, wcs = cut(data, wcs, scale=scale)
|
@@ -96,13 +94,13 @@ def cut_n_predict(data, wcs, scale):
|
|
96 |
|
97 |
return y_pred, wcs
|
98 |
|
99 |
-
# Define function to
|
100 |
@st.cache
|
101 |
def decompose_cavity(pred, th2=0.7, amin=10):
|
102 |
X, Y = pred.nonzero()
|
103 |
data = np.array([X,Y]).reshape(2, -1)
|
104 |
|
105 |
-
# DBSCAN
|
106 |
try: clusters = DBSCAN(eps=1.0, min_samples=3).fit(data.T).labels_
|
107 |
except: clusters = []
|
108 |
|
@@ -115,14 +113,16 @@ def decompose_cavity(pred, th2=0.7, amin=10):
|
|
115 |
xi, yi = X[b], Y[b]
|
116 |
img[xi, yi] = pred[xi, yi]
|
117 |
|
118 |
-
#
|
119 |
if not (img > th2).any(): continue
|
120 |
|
121 |
-
#
|
122 |
if np.sum(img) <= amin: continue
|
123 |
|
124 |
cavities.append(img)
|
125 |
|
|
|
|
|
126 |
ccd = CCDData(pred, unit="adu", wcs=wcs)
|
127 |
ccd.write(f"predictions/predicted.fits", overwrite=True)
|
128 |
image_decomposed = np.zeros((128,128))
|
@@ -133,24 +133,14 @@ def decompose_cavity(pred, th2=0.7, amin=10):
|
|
133 |
|
134 |
return image_decomposed
|
135 |
|
136 |
-
#
|
137 |
-
|
138 |
-
# shutil.make_archive("predictions.zip", 'zip', "predictions")
|
139 |
-
# with open('predictions.zip', 'rb') as f:
|
140 |
-
# res = f.read()
|
141 |
-
# return res
|
142 |
-
|
143 |
bordersize = 0.6
|
144 |
_, col, _ = st.columns([bordersize, 3, bordersize])
|
145 |
|
146 |
-
# if os.path.exists("pred.npy"): os.system("rm pred.npy")
|
147 |
-
# os.system("rm -r predictions")
|
148 |
-
# os.system("rm predictions.zip Views")
|
149 |
-
os.system("mkdir -p predictions")
|
150 |
-
|
151 |
with col:
|
152 |
-
|
153 |
-
|
154 |
st.markdown("Cavity Detection Tool (CADET) is a machine learning pipeline trained to detect X-ray cavities from noisy Chandra images of early-type galaxies.")
|
155 |
st.markdown("To use this tool: upload your image, select the scale of interest, and make a prediction!")
|
156 |
st.markdown("Input images should be centred at the centre of the galaxy and point sources should be filled with surrounding background ([dmfilth](https://cxc.cfa.harvard.edu/ciao/ahelp/dmfilth.html)).")
|
@@ -170,61 +160,46 @@ if uploaded_file is not None:
|
|
170 |
col1.subheader("Input image")
|
171 |
col3.subheader("Prediction")
|
172 |
col5.subheader("Decomposed")
|
173 |
-
col6.subheader("")
|
174 |
|
175 |
with col1:
|
176 |
-
|
177 |
# st.markdown("""<style>[data-baseweb="select"] {margin-top: -26px;}</style>""", unsafe_allow_html=True)
|
178 |
max_scale = int(data.shape[0] // 128)
|
179 |
scale = st.selectbox('Scale:',[f"{(i+1)*128}x{(i+1)*128}" for i in range(max_scale)], label_visibility="hidden")
|
180 |
scale = int(scale.split("x")[0]) // 128
|
181 |
-
|
182 |
-
with col3:
|
183 |
-
detect = st.button('Detect')
|
184 |
|
185 |
-
|
186 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
187 |
|
188 |
# Make two columns for plots
|
189 |
_, colA, colB, colC, _ = st.columns([bordersize,1,1,1,bordersize])
|
190 |
|
191 |
image = np.log10(data+1)
|
192 |
plot_image(image, scale)
|
193 |
-
|
194 |
-
|
195 |
-
with col4:
|
196 |
-
st.markdown("""<style>[data-baseweb="select"] {margin-top: -36px;}</style>""", unsafe_allow_html=True)
|
197 |
-
threshold = st.slider("", 0.0, 1.0, 0.0, 0.05, label_visibility="hidden")
|
198 |
|
199 |
if detect or threshold:
|
200 |
y_pred, wcs = cut_n_predict(data, wcs, scale)
|
201 |
|
202 |
-
#
|
203 |
-
|
204 |
-
# try: y_pred = np.load("pred.npy")
|
205 |
# except: y_pred = np.zeros((128,128))
|
206 |
-
try: _ = y_pred
|
207 |
-
except: y_pred = np.zeros((128,128))
|
208 |
y_pred_th = np.where(y_pred > threshold, y_pred, 0)
|
209 |
-
# np.save("thresh.npy", y_pred)
|
210 |
|
211 |
plot_prediction(y_pred_th)
|
212 |
|
213 |
if decompose:
|
214 |
-
# y_pred = np.load("thresh.npy")
|
215 |
|
216 |
image_decomposed = decompose_cavity(y_pred_th)
|
217 |
|
218 |
-
#
|
219 |
-
#
|
220 |
-
# image_decomposed = np.zeros((128,128))
|
221 |
-
# for i, cav in enumerate(cavs):
|
222 |
-
# ccd = CCDData(cav, unit="adu", wcs=wcs)
|
223 |
-
# ccd.write(f"predictions/predicted_{i+1}.fits", overwrite=True)
|
224 |
-
# image_decomposed += (i+1) * np.where(cav > 0, 1, 0)
|
225 |
-
|
226 |
-
try: _ = image_decomposed
|
227 |
-
except: image_decomposed = np.zeros((128,128))
|
228 |
plot_decomposed(image_decomposed)
|
229 |
|
230 |
with col6:
|
@@ -235,4 +210,4 @@ if uploaded_file is not None:
|
|
235 |
st.markdown("")
|
236 |
# st.markdown("""<style>[data-baseweb="select"] {margin-top: 16px;}</style>""", unsafe_allow_html=True)
|
237 |
fname = uploaded_file.name.strip(".fits")
|
238 |
-
download = st.download_button(label="Download", data=res, file_name=f'
|
|
|
24 |
# Streamlit
|
25 |
import streamlit as st
|
26 |
st.set_option('deprecation.showPyplotGlobalUse', False)
|
|
|
|
|
27 |
|
28 |
# Define function to plot the uploaded image
|
29 |
def plot_image(image, scale):
|
|
|
56 |
plt.axis('off')
|
57 |
with colC: st.pyplot()
|
58 |
|
59 |
+
# Define function to cut input image and rebin it to 128x128 pixels
|
60 |
def cut(data0, wcs0, scale=1):
|
61 |
shape = data0.shape[0]
|
62 |
x0 = shape / 2
|
|
|
79 |
|
80 |
return data, wcs
|
81 |
|
82 |
+
# Define function to apply cutting and produce a prediction
|
83 |
@st.cache
|
84 |
def cut_n_predict(data, wcs, scale):
|
85 |
data, wcs = cut(data, wcs, scale=scale)
|
|
|
94 |
|
95 |
return y_pred, wcs
|
96 |
|
97 |
+
# Define function to decompose prediction into individual cavities
|
98 |
@st.cache
|
99 |
def decompose_cavity(pred, th2=0.7, amin=10):
|
100 |
X, Y = pred.nonzero()
|
101 |
data = np.array([X,Y]).reshape(2, -1)
|
102 |
|
103 |
+
# DBSCAN clustering
|
104 |
try: clusters = DBSCAN(eps=1.0, min_samples=3).fit(data.T).labels_
|
105 |
except: clusters = []
|
106 |
|
|
|
113 |
xi, yi = X[b], Y[b]
|
114 |
img[xi, yi] = pred[xi, yi]
|
115 |
|
116 |
+
# Thresholding #2
|
117 |
if not (img > th2).any(): continue
|
118 |
|
119 |
+
# Minimal area
|
120 |
if np.sum(img) <= amin: continue
|
121 |
|
122 |
cavities.append(img)
|
123 |
|
124 |
+
# Save raw and decomposed predictions to predictions folder
|
125 |
+
os.system("mkdir -p predictions")
|
126 |
ccd = CCDData(pred, unit="adu", wcs=wcs)
|
127 |
ccd.write(f"predictions/predicted.fits", overwrite=True)
|
128 |
image_decomposed = np.zeros((128,128))
|
|
|
133 |
|
134 |
return image_decomposed
|
135 |
|
136 |
+
# Use wide layout and create columns
|
137 |
+
st.set_page_config(page_title="Cavity Detection Tool", layout="wide")
|
|
|
|
|
|
|
|
|
|
|
138 |
bordersize = 0.6
|
139 |
_, col, _ = st.columns([bordersize, 3, bordersize])
|
140 |
|
|
|
|
|
|
|
|
|
|
|
141 |
with col:
|
142 |
+
# Create heading and description
|
143 |
+
st.markdown("# Cavity Detection Tool")
|
144 |
st.markdown("Cavity Detection Tool (CADET) is a machine learning pipeline trained to detect X-ray cavities from noisy Chandra images of early-type galaxies.")
|
145 |
st.markdown("To use this tool: upload your image, select the scale of interest, and make a prediction!")
|
146 |
st.markdown("Input images should be centred at the centre of the galaxy and point sources should be filled with surrounding background ([dmfilth](https://cxc.cfa.harvard.edu/ciao/ahelp/dmfilth.html)).")
|
|
|
160 |
col1.subheader("Input image")
|
161 |
col3.subheader("Prediction")
|
162 |
col5.subheader("Decomposed")
|
163 |
+
# col6.subheader("")
|
164 |
|
165 |
with col1:
|
|
|
166 |
# st.markdown("""<style>[data-baseweb="select"] {margin-top: -26px;}</style>""", unsafe_allow_html=True)
|
167 |
max_scale = int(data.shape[0] // 128)
|
168 |
scale = st.selectbox('Scale:',[f"{(i+1)*128}x{(i+1)*128}" for i in range(max_scale)], label_visibility="hidden")
|
169 |
scale = int(scale.split("x")[0]) // 128
|
|
|
|
|
|
|
170 |
|
171 |
+
# Detect button
|
172 |
+
with col3: detect = st.button('Detect')
|
173 |
+
|
174 |
+
# Threshold slider
|
175 |
+
with col4:
|
176 |
+
# st.markdown("""<style>[data-baseweb="select"] {margin-top: -36px;}</style>""", unsafe_allow_html=True)
|
177 |
+
threshold = st.slider("Threshold", 0.0, 1.0, 0.0, 0.05) #, label_visibility="hidden")
|
178 |
+
|
179 |
+
# Decompose button
|
180 |
+
with col5: decompose = st.button('Decompose')
|
181 |
|
182 |
# Make two columns for plots
|
183 |
_, colA, colB, colC, _ = st.columns([bordersize,1,1,1,bordersize])
|
184 |
|
185 |
image = np.log10(data+1)
|
186 |
plot_image(image, scale)
|
|
|
|
|
|
|
|
|
|
|
187 |
|
188 |
if detect or threshold:
|
189 |
y_pred, wcs = cut_n_predict(data, wcs, scale)
|
190 |
|
191 |
+
# try: _ = y_pred
|
|
|
|
|
192 |
# except: y_pred = np.zeros((128,128))
|
|
|
|
|
193 |
y_pred_th = np.where(y_pred > threshold, y_pred, 0)
|
|
|
194 |
|
195 |
plot_prediction(y_pred_th)
|
196 |
|
197 |
if decompose:
|
|
|
198 |
|
199 |
image_decomposed = decompose_cavity(y_pred_th)
|
200 |
|
201 |
+
# try: _ = image_decomposed
|
202 |
+
# except: image_decomposed = np.zeros((128,128))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
203 |
plot_decomposed(image_decomposed)
|
204 |
|
205 |
with col6:
|
|
|
210 |
st.markdown("")
|
211 |
# st.markdown("""<style>[data-baseweb="select"] {margin-top: 16px;}</style>""", unsafe_allow_html=True)
|
212 |
fname = uploaded_file.name.strip(".fits")
|
213 |
+
download = st.download_button(label="Download", data=res, file_name=f'{fname}_{int(scale*128)}.zip', mime="application/octet-stream")
|