Update app.py
Browse files
app.py
CHANGED
@@ -4,6 +4,7 @@ import matplotlib.pyplot as plt
|
|
4 |
from torchvision import transforms
|
5 |
from PIL import Image
|
6 |
import torch.nn as nn
|
|
|
7 |
|
8 |
from PathDino import get_pathDino_model
|
9 |
|
@@ -82,13 +83,19 @@ def visualize_attention_ViT(model, img, patch_size=16):
|
|
82 |
return attention_list
|
83 |
|
84 |
# Define the function to generate activation maps
|
85 |
-
def generate_activation_maps(image):
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
preprocess = transforms.Compose([
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
image_tensor = preprocess(image)
|
93 |
img = image_tensor.unsqueeze(0).to(device)
|
94 |
# Generate activation maps
|
@@ -106,15 +113,16 @@ uploaded_image = st.file_uploader("Upload an image", type=["jpg", "png", "jpeg"]
|
|
106 |
|
107 |
if uploaded_image is not None:
|
108 |
columns = st.columns(3)
|
109 |
-
columns[1].image(uploaded_image, caption="Uploaded Image", width=
|
110 |
|
111 |
# Load the image and apply preprocessing
|
112 |
uploaded_image = Image.open(uploaded_image).convert('RGB')
|
113 |
attention_list = generate_activation_maps(uploaded_image)
|
114 |
print(len(attention_list))
|
115 |
st.subheader(f"Attention Maps of the input image")
|
116 |
-
columns = st.columns(
|
117 |
-
columns2 = st.columns(
|
|
|
118 |
for index, col in enumerate(columns):
|
119 |
# Create a plot
|
120 |
plt.plot(512, 512)
|
@@ -133,7 +141,25 @@ if uploaded_image is not None:
|
|
133 |
|
134 |
for index, col in enumerate(columns2):
|
135 |
|
136 |
-
index = index +
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
137 |
# Create a plot
|
138 |
plt.plot(512, 512)
|
139 |
|
|
|
4 |
from torchvision import transforms
|
5 |
from PIL import Image
|
6 |
import torch.nn as nn
|
7 |
+
import numpy as np
|
8 |
|
9 |
from PathDino import get_pathDino_model
|
10 |
|
|
|
83 |
return attention_list
|
84 |
|
85 |
# Define the function to generate activation maps
|
86 |
+
def generate_activation_maps(image, patch_size=16):
|
87 |
+
# Convert the image to a NumPy array
|
88 |
+
img = np.array(image)
|
89 |
+
# make the image divisible by the patch size
|
90 |
+
w, h = img.shape[1] - img.shape[0] % patch_size, img.shape[1] - img.shape[1] % patch_size
|
91 |
+
min_size = min(w, h)
|
92 |
+
print("Image shape:", img.shape)
|
93 |
preprocess = transforms.Compose([
|
94 |
+
transforms.Resize((img.shape[0], img.shape[1])),
|
95 |
+
transforms.CenterCrop(min_size),
|
96 |
+
transforms.ToTensor(),
|
97 |
+
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) # Normalize the tensors
|
98 |
+
])
|
99 |
image_tensor = preprocess(image)
|
100 |
img = image_tensor.unsqueeze(0).to(device)
|
101 |
# Generate activation maps
|
|
|
113 |
|
114 |
if uploaded_image is not None:
|
115 |
columns = st.columns(3)
|
116 |
+
columns[1].image(uploaded_image, caption="Uploaded Image", width=400)
|
117 |
|
118 |
# Load the image and apply preprocessing
|
119 |
uploaded_image = Image.open(uploaded_image).convert('RGB')
|
120 |
attention_list = generate_activation_maps(uploaded_image)
|
121 |
print(len(attention_list))
|
122 |
st.subheader(f"Attention Maps of the input image")
|
123 |
+
columns = st.columns(2)
|
124 |
+
columns2 = st.columns(2)
|
125 |
+
columns3 = st.columns(2)
|
126 |
for index, col in enumerate(columns):
|
127 |
# Create a plot
|
128 |
plt.plot(512, 512)
|
|
|
141 |
|
142 |
for index, col in enumerate(columns2):
|
143 |
|
144 |
+
index = index + 2
|
145 |
+
# Create a plot
|
146 |
+
plt.plot(512, 512)
|
147 |
+
|
148 |
+
# Remove x and y axis labels
|
149 |
+
plt.xticks([]) # Hide x-axis ticks and labels
|
150 |
+
plt.yticks([]) # Hide y-axis ticks and labels
|
151 |
+
|
152 |
+
# Alternatively, if you only want to hide the labels and keep the ticks:
|
153 |
+
plt.gca().axes.get_xaxis().set_visible(False)
|
154 |
+
plt.gca().axes.get_yaxis().set_visible(False)
|
155 |
+
|
156 |
+
plt.imshow(attention_list[index])
|
157 |
+
col.pyplot(plt)
|
158 |
+
plt.close()
|
159 |
+
|
160 |
+
for index, col in enumerate(columns3):
|
161 |
+
|
162 |
+
index = index + 4
|
163 |
# Create a plot
|
164 |
plt.plot(512, 512)
|
165 |
|