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