Update New_file.txt
Browse files- New_file.txt +52 -110
New_file.txt
CHANGED
@@ -1,112 +1,54 @@
|
|
1 |
-
import
|
2 |
-
import torchvision.transforms as transforms
|
3 |
-
from PIL import Image
|
4 |
import numpy as np
|
5 |
-
from sklearn.metrics.pairwise import cosine_similarity
|
6 |
-
from torchvision.models import resnet50
|
7 |
-
from torchvision.datasets import ImageFolder
|
8 |
-
from torch.utils.data import DataLoader
|
9 |
-
|
10 |
-
# Load a pre-trained ResNet-50 model
|
11 |
-
model = resnet50(pretrained=True)
|
12 |
-
model.eval()
|
13 |
-
|
14 |
-
# Define a function to preprocess images
|
15 |
-
def preprocess_image(image_path):
|
16 |
-
transform = transforms.Compose([
|
17 |
-
transforms.Resize((224, 224)),
|
18 |
-
transforms.ToTensor(),
|
19 |
-
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
|
20 |
-
])
|
21 |
-
image = Image.open(image_path)
|
22 |
-
image = transform(image).unsqueeze(0) # Add a batch dimension
|
23 |
-
return image
|
24 |
-
|
25 |
-
# Load your ideal subset of images
|
26 |
-
ideal_image_paths = ["/content/trunck.jpg", "t4.jpg"] # Replace with your ideal image file paths
|
27 |
-
ideal_embeddings = []
|
28 |
-
|
29 |
-
for image_path in ideal_image_paths:
|
30 |
-
image = preprocess_image(image_path)
|
31 |
-
with torch.no_grad():
|
32 |
-
embedding = model(image).squeeze().numpy()
|
33 |
-
ideal_embeddings.append(embedding)
|
34 |
-
|
35 |
-
# Load a set of candidate images
|
36 |
-
candidate_image_paths = ["/content/trunck2.jpg", "t3.jpg", "car.jpg",] # Replace with your candidate image file paths
|
37 |
-
candidate_embeddings = []
|
38 |
-
|
39 |
-
for image_path in candidate_image_paths:
|
40 |
-
image = preprocess_image(image_path)
|
41 |
-
with torch.no_grad():
|
42 |
-
embedding = model(image).squeeze().numpy()
|
43 |
-
candidate_embeddings.append(embedding)
|
44 |
-
|
45 |
-
# Calculate similarities between ideal and candidate images using cosine similarity
|
46 |
-
similarities = cosine_similarity(ideal_embeddings, candidate_embeddings)
|
47 |
-
|
48 |
-
# Print the similarity matrix
|
49 |
-
print(similarities)
|
50 |
-
|
51 |
-
|
52 |
-
import torch
|
53 |
-
from transformers import SwinTransformer, SwinTransformerImageProcessor
|
54 |
-
import torchvision.transforms as transforms
|
55 |
-
from PIL import Image
|
56 |
-
import numpy as np
|
57 |
-
from sklearn.metrics.pairwise import cosine_similarity
|
58 |
-
|
59 |
-
# Load the pretrained Swin Transformer model and image processor
|
60 |
-
model_name = "microsoft/Swin-Transformer-base-patch4-in22k"
|
61 |
-
model = SwinTransformer.from_pretrained(model_name)
|
62 |
-
processor = SwinTransformerImageProcessor.from_pretrained(model_name)
|
63 |
-
|
64 |
-
# Define a function to preprocess images
|
65 |
-
def preprocess_image(image_path):
|
66 |
-
transform = transforms.Compose([
|
67 |
-
transforms.Resize((224, 224)),
|
68 |
-
transforms.ToTensor(),
|
69 |
-
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
|
70 |
-
])
|
71 |
-
image = Image.open(image_path)
|
72 |
-
inputs = processor(images=image, return_tensors="pt")
|
73 |
-
return inputs
|
74 |
-
|
75 |
-
# Load your ideal and candidate subsets of images
|
76 |
-
ideal_image_paths = ["ideal_image1.jpg", "ideal_image2.jpg", "ideal_image3.jpg"] # Replace with your ideal image file paths
|
77 |
-
candidate_image_paths = ["candidate_image1.jpg", "candidate_image2.jpg", "candidate_image3.jpg"] # Replace with your candidate image file paths
|
78 |
-
|
79 |
-
# Calculate cosine similarities between ideal and candidate images
|
80 |
-
similarities = []
|
81 |
-
|
82 |
-
for ideal_path in ideal_image_paths:
|
83 |
-
ideal_embedding = None
|
84 |
-
inputs_ideal = preprocess_image(ideal_path)
|
85 |
-
with torch.no_grad():
|
86 |
-
output_ideal = model(**inputs_ideal)
|
87 |
-
ideal_embedding = output_ideal['pixel_values'][0].cpu().numpy()
|
88 |
-
|
89 |
-
for candidate_path in candidate_image_paths:
|
90 |
-
candidate_embedding = None
|
91 |
-
inputs_candidate = preprocess_image(candidate_path)
|
92 |
-
with torch.no_grad():
|
93 |
-
output_candidate = model(**inputs_candidate)
|
94 |
-
candidate_embedding = output_candidate['pixel_values'][0].cpu().numpy()
|
95 |
-
|
96 |
-
# Calculate cosine similarity between ideal and candidate embeddings
|
97 |
-
similarity = cosine_similarity([ideal_embedding], [candidate_embedding])[0][0]
|
98 |
-
similarities.append((ideal_path, candidate_path, similarity))
|
99 |
-
|
100 |
-
# Set a similarity threshold (e.g., 0.7)
|
101 |
-
threshold = 0.7
|
102 |
-
|
103 |
-
# Find similar image pairs based on the threshold
|
104 |
-
similar_pairs = []
|
105 |
-
for ideal_path, candidate_path, similarity in similarities:
|
106 |
-
if similarity > threshold:
|
107 |
-
similar_pairs.append((ideal_path, candidate_path))
|
108 |
-
|
109 |
-
# Print similar image pairs
|
110 |
-
for pair in similar_pairs:
|
111 |
-
print(f"Similar images: {pair[0]} and {pair[1]}")
|
112 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import cv2
|
|
|
|
|
2 |
import numpy as np
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
|
4 |
+
# Load the images
|
5 |
+
image1_path = 'path_to_your_first_image.jpg'
|
6 |
+
image2_path = 'path_to_your_second_image.jpg'
|
7 |
+
image1 = cv2.imread(image1_path)
|
8 |
+
image2 = cv2.imread(image2_path)
|
9 |
+
|
10 |
+
# Resize images to the same height for concatenation
|
11 |
+
height1, width1, _ = image1.shape
|
12 |
+
height2, width2, _ = image2.shape
|
13 |
+
|
14 |
+
# Define the desired height for both images (e.g., height of the first image)
|
15 |
+
desired_height = height1
|
16 |
+
|
17 |
+
# Resize images
|
18 |
+
image1_resized = cv2.resize(image1, (width1, desired_height))
|
19 |
+
image2_resized = cv2.resize(image2, (width2, desired_height))
|
20 |
+
|
21 |
+
# Combine images side by side
|
22 |
+
combined_image = np.hstack((image1_resized, image2_resized))
|
23 |
+
|
24 |
+
# Add labels to the top of each image
|
25 |
+
label1 = 'Image 1'
|
26 |
+
label2 = 'Image 2'
|
27 |
+
font = cv2.FONT_HERSHEY_SIMPLEX
|
28 |
+
font_scale = 1
|
29 |
+
color = (255, 255, 255) # White color for the text
|
30 |
+
thickness = 2
|
31 |
+
|
32 |
+
# Calculate the position for the labels
|
33 |
+
label1_size = cv2.getTextSize(label1, font, font_scale, thickness)[0]
|
34 |
+
label2_size = cv2.getTextSize(label2, font, font_scale, thickness)[0]
|
35 |
+
|
36 |
+
# Position for label1
|
37 |
+
x1 = 10
|
38 |
+
y1 = label1_size[1] + 10
|
39 |
+
|
40 |
+
# Position for label2
|
41 |
+
x2 = width1 + 10
|
42 |
+
y2 = label2_size[1] + 10
|
43 |
+
|
44 |
+
# Add labels to the combined image
|
45 |
+
cv2.putText(combined_image, label1, (x1, y1), font, font_scale, color, thickness)
|
46 |
+
cv2.putText(combined_image, label2, (x2, y2), font, font_scale, color, thickness)
|
47 |
+
|
48 |
+
# Display the combined image
|
49 |
+
cv2.imshow('Combined Image', combined_image)
|
50 |
+
cv2.waitKey(0)
|
51 |
+
cv2.destroyAllWindows()
|
52 |
+
|
53 |
+
# Save the combined image
|
54 |
+
cv2.imwrite('combined_image.jpg', combined_image)
|