Upload 18 files

.gitattributes

@@ -32,3 +32,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+data/results.csv filter=lfs diff=lfs merge=lfs -text

data/results.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cf240f2c5d6c4e9cd0d1acef589e36225a3622ad93bb6e1fe7f828a4c9fb31d
+size 13350817

embeddings/emb_images_30k.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:154dab0b8ad9455c4df7d819b471987f23d46c63633fc76a480bbbba9e2c20a8
+size 32545920

embeddings/emb_images_5000.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec835458c3f03a01564e3a2c0bbc2cc5326f0f52fae2eb6edd02ad0d0eb1734e
+size 10240128

funcs/fiass_similaruty.py

@@ -0,0 +1,34 @@
+import torch
+import numpy as np
+import clip
+import torch.nn.functional as F
+import faiss
+
+device = 'cpu'
+model_path = "weights/ViT-B-32.pt"
+
+model, preprocess = clip.load('ViT-B/32', device)
+
+def load_embeddings(path_to_emb_file):
+    features = np.load(path_to_emb_file)
+    features = torch.from_numpy(features)
+    features = features.squeeze(1)
+    features = F.normalize(features, p=2, dim=-1)
+    return features
+
+def encode_text(query):
+    text = clip.tokenize([query]).to(device)
+    text_features = model.encode_text(text).to("cpu")
+    text_features= F.normalize(text_features, p=2, dim=-1)
+    text_features = text_features.to("cpu").detach().numpy()
+    return text_features
+
+
+def find_matches_fiass(image_embeddings, query, image_filenames, n=5):
+    features = image_embeddings
+    index = faiss.IndexFlatL2(features.shape[1])
+    index.add(features)
+    text_features = encode_text(query)
+    _, I = index.search(text_features, n)
+    matches = [image_filenames[idx] for idx in I.squeeze(0)]
+    return matches 

funcs/get_embeddings.py

@@ -0,0 +1,34 @@
+import clip
+import os
+from PIL import Image
+import numpy as np
+import torch
+
+device = 'cpu'
+model_path = "weights/ViT-B-32.pt"
+
+model, preprocess = clip.load('ViT-B/32', device)
+
+
+def get_emb(image_folder):
+    image_folder = image_folder
+    image_paths = []
+    for filename in os.listdir(image_folder):
+            if filename.endswith(".jpg") or filename.endswith(".jpeg") or filename.endswith(".png"):
+                    image_path = os.path.join(image_folder, filename)
+                    image_paths.append(image_path)
+            paths = image_paths
+    images = [Image.open(path) for path in image_paths]
+    with torch.no_grad():
+        features = []
+        for image in images:
+            image_tensor = preprocess(image).unsqueeze(0).to(device)
+            feature = model.encode_image(image_tensor)
+            features.append(feature.detach().cpu().numpy())
+        features = np.array(features)
+    np.save("emb_images.npy", features)
+    return features, paths
+        
+    
+
+            

funcs/get_similarity.py

@@ -0,0 +1,52 @@
+import streamlit as st
+import torch
+import torchvision.transforms as transforms
+from PIL import Image
+import os
+import clip
+import numpy as np
+import torch.nn.functional as F
+import matplotlib.pyplot as plt
+
+device = 'cpu'
+model_path = "weights/ViT-B-32.pt"
+
+model, preprocess = clip.load('ViT-B/32', device)
+
+def get_similarity_score(text_query, image_features):
+    text_tokens = clip.tokenize([text_query]).to(device)
+    with torch.no_grad():
+        text_features = model.encode_text(text_tokens).squeeze(0)
+        text_features= F.normalize(text_features, p=2, dim=-1)
+        similarity_score = text_features @ image_features.T * 100.0
+        similarity_score = similarity_score.squeeze(0)
+    return similarity_score
+
+def create_filelist(path_to_imagefolder):
+        image_folder = path_to_imagefolder
+        image_paths = []
+        for filename in os.listdir(image_folder):
+                if filename.endswith(".jpg") or filename.endswith(".jpeg") or filename.endswith(".png"):
+                        image_path = os.path.join(image_folder, filename)
+                        image_paths.append(image_path)
+                file_paths = image_paths
+        return file_paths
+
+def load_embeddings(path_to_emb_file):
+    features = np.load(path_to_emb_file)
+    features = torch.from_numpy(features)
+    return features
+
+
+def find_matches(image_embeddings, query, image_filenames, n=6):
+    text_query = query
+    features = image_embeddings
+    similarity_scores = []
+    for emb in features:
+        emb /= emb.norm(dim=-1, keepdim=True)
+        similarity_score = get_similarity_score(text_query, emb)
+        similarity_scores.append(similarity_score)
+    similarity_scores = torch.stack(similarity_scores)
+    values, indices = torch.topk(similarity_scores.squeeze(0), 6)
+    matches = [image_filenames[idx] for idx in indices]
+    return matches