import streamlit as st import open_clip import torch import requests from PIL import Image from io import BytesIO import time import json import numpy as np import cv2 import chromadb from transformers import SegformerImageProcessor, AutoModelForSemanticSegmentation import torch.nn as nn import matplotlib.pyplot as plt # Load CLIP model and tokenizer @st.cache_resource def load_clip_model(): model, preprocess_train, preprocess_val = open_clip.create_model_and_transforms('hf-hub:Marqo/marqo-fashionSigLIP') tokenizer = open_clip.get_tokenizer('hf-hub:Marqo/marqo-fashionSigLIP') device = torch.device("cuda" if torch.cuda.is_available() else "cpu") model.to(device) return model, preprocess_val, tokenizer, device clip_model, preprocess_val, tokenizer, device = load_clip_model() # Load SegFormer model @st.cache_resource def load_segformer_model(): processor = SegformerImageProcessor.from_pretrained("sayeed99/segformer-b3-fashion") model = AutoModelForSemanticSegmentation.from_pretrained("sayeed99/segformer-b3-fashion") return model, processor segformer_model, segformer_processor = load_segformer_model() # Load ChromaDB @st.cache_resource def load_chromadb(): client = chromadb.PersistentClient(path="./clothesDB") collection = client.get_collection(name="clothes_items_ver3") return collection collection = load_chromadb() # Helper functions def load_image_from_url(url, max_retries=3): for attempt in range(max_retries): try: response = requests.get(url, timeout=10) response.raise_for_status() img = Image.open(BytesIO(response.content)).convert('RGB') return img except (requests.RequestException, Image.UnidentifiedImageError) as e: if attempt < max_retries - 1: time.sleep(1) else: return None def get_image_embedding(image): image_tensor = preprocess_val(image).unsqueeze(0).to(device) with torch.no_grad(): image_features = clip_model.encode_image(image_tensor) image_features /= image_features.norm(dim=-1, keepdim=True) return image_features.cpu().numpy() def get_text_embedding(text): text_tokens = tokenizer([text]).to(device) with torch.no_grad(): text_features = clip_model.encode_text(text_tokens) text_features /= text_features.norm(dim=-1, keepdim=True) return text_features.cpu().numpy() def find_similar_images(query_embedding, collection, top_k=5): database_embeddings = np.array(collection.get(include=['embeddings'])['embeddings']) similarities = np.dot(database_embeddings, query_embedding.T).squeeze() top_indices = np.argsort(similarities)[::-1][:top_k] all_data = collection.get(include=['metadatas'])['metadatas'] results = [ {'info': all_data[idx], 'similarity': similarities[idx]} for idx in top_indices ] return results def segment_clothing(image): inputs = segformer_processor(images=image, return_tensors="pt") outputs = segformer_model(**inputs) logits = outputs.logits upsampled_logits = nn.functional.interpolate( logits, size=image.size[::-1], mode="bilinear", align_corners=False, ) pred_seg = upsampled_logits.argmax(dim=1)[0].cpu().numpy() categories = segformer_model.config.id2label segmented_items = [] for category_id, category_name in categories.items(): if category_id in pred_seg: mask = (pred_seg == category_id).astype(np.uint8) contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) if contours: x, y, w, h = cv2.boundingRect(max(contours, key=cv2.contourArea)) segmented_items.append({ 'category': category_name, 'bbox': [x, y, x+w, y+h], 'mask': mask }) return segmented_items, pred_seg, categories def visualize_segmentation(pred_seg, categories): plt.figure(figsize=(10, 10)) plt.imshow(pred_seg, cmap='jet') plt.colorbar(label='Category ID') plt.title("Segmentation Result") plt.axis('off') # Add legend unique_classes = np.unique(pred_seg) legend_elements = [plt.Rectangle((0,0),1,1, color=plt.cm.jet(category_id/len(categories)), label=f"{category_id}: {categories[category_id]}") for category_id in unique_classes if category_id in categories] plt.legend(handles=legend_elements, loc='center left', bbox_to_anchor=(1, 0.5)) return plt def crop_image(image, bbox): return image.crop((bbox[0], bbox[1], bbox[2], bbox[3])) # Streamlit app st.title("Advanced Fashion Search App") # Initialize session state if 'step' not in st.session_state: st.session_state.step = 'input' if 'query_image_url' not in st.session_state: st.session_state.query_image_url = '' if 'segmentations' not in st.session_state: st.session_state.segmentations = [] if 'selected_category' not in st.session_state: st.session_state.selected_category = None # Step-by-step processing if st.session_state.step == 'input': st.session_state.query_image_url = st.text_input("Enter image URL:", st.session_state.query_image_url) if st.button("Segment Clothing"): if st.session_state.query_image_url: query_image = load_image_from_url(st.session_state.query_image_url) if query_image is not None: st.session_state.query_image = query_image st.session_state.segmentations, st.session_state.pred_seg, st.session_state.categories = segment_clothing(query_image) if st.session_state.segmentations: st.session_state.step = 'select_category' else: st.warning("No clothing items segmented in the image.") else: st.error("Failed to load the image. Please try another URL.") else: st.warning("Please enter an image URL.") elif st.session_state.step == 'select_category': col1, col2 = st.columns(2) with col1: st.image(st.session_state.query_image, caption="Original Image", use_column_width=True) with col2: seg_fig = visualize_segmentation(st.session_state.pred_seg, st.session_state.categories) st.pyplot(seg_fig) plt.close(seg_fig) # Prevent memory leaks st.subheader("Segmented Clothing Items:") for segmentation in st.session_state.segmentations: col1, col2 = st.columns([1, 3]) with col1: st.write(f"{segmentation['category']}") with col2: cropped_image = crop_image(st.session_state.query_image, segmentation['bbox']) st.image(cropped_image, caption=segmentation['category'], use_column_width=True) options = [s['category'] for s in st.session_state.segmentations] selected_option = st.selectbox("Select a category to search:", options) if st.button("Search Similar Items"): st.session_state.selected_category = selected_option st.session_state.step = 'show_results' elif st.session_state.step == 'show_results': st.image(st.session_state.query_image, caption="Query Image", use_column_width=True) selected_segmentation = next(s for s in st.session_state.segmentations if s['category'] == st.session_state.selected_category) cropped_image = crop_image(st.session_state.query_image, selected_segmentation['bbox']) st.image(cropped_image, caption="Cropped Image", use_column_width=True) query_embedding = get_image_embedding(cropped_image) similar_images = find_similar_images(query_embedding, collection) st.subheader("Similar Items:") for img in similar_images: col1, col2 = st.columns(2) with col1: st.image(img['info']['image_url'], use_column_width=True) with col2: st.write(f"Name: {img['info']['name']}") st.write(f"Brand: {img['info']['brand']}") category = img['info'].get('category') if category: st.write(f"Category: {category}") st.write(f"Price: {img['info']['price']}") st.write(f"Discount: {img['info']['discount']}%") st.write(f"Similarity: {img['similarity']:.2f}") if st.button("Start New Search"): st.session_state.step = 'input' st.session_state.query_image_url = '' st.session_state.segmentations = [] st.session_state.selected_category = None else: # Text search query_text = st.text_input("Enter search text:") if st.button("Search by Text"): if query_text: text_embedding = get_text_embedding(query_text) similar_images = find_similar_images(text_embedding, collection) st.subheader("Similar Items:") for img in similar_images: col1, col2 = st.columns(2) with col1: st.image(img['info']['image_url'], use_column_width=True) with col2: st.write(f"Name: {img['info']['name']}") st.write(f"Brand: {img['info']['brand']}") category = img['info'].get('category') if category: st.write(f"Category: {category}") st.write(f"Price: {img['info']['price']}") st.write(f"Discount: {img['info']['discount']}%") st.write(f"Similarity: {img['similarity']:.2f}") else: st.warning("Please enter a search text.") # Display ChromaDB vacuum message st.sidebar.warning("If you've upgraded ChromaDB from a version below 0.6, you may benefit from vacuuming your database. Run 'chromadb utils vacuum --help' for more information.")