Spaces:
Running
Running
| import streamlit as st | |
| import pandas as pd | |
| from plip_support import embed_text | |
| import numpy as np | |
| from PIL import Image | |
| import requests | |
| import pickle | |
| import tokenizers | |
| from io import BytesIO | |
| import torch | |
| from transformers import ( | |
| VisionTextDualEncoderModel, | |
| AutoFeatureExtractor, | |
| AutoTokenizer, | |
| CLIPModel, | |
| AutoProcessor | |
| ) | |
| import streamlit.components.v1 as components | |
| def embed_texts(model, texts, processor): | |
| inputs = processor(text=texts, padding="longest") | |
| input_ids = torch.tensor(inputs["input_ids"]) | |
| attention_mask = torch.tensor(inputs["attention_mask"]) | |
| with torch.no_grad(): | |
| embeddings = model.get_text_features( | |
| input_ids=input_ids, attention_mask=attention_mask | |
| ) | |
| return embeddings | |
| def load_embeddings(embeddings_path): | |
| print("loading embeddings") | |
| return np.load(embeddings_path) | |
| def load_path_clip(): | |
| model = CLIPModel.from_pretrained("vinid/plip") | |
| processor = AutoProcessor.from_pretrained("vinid/plip") | |
| return model, processor | |
| def init(): | |
| with open('data/twitter.asset', 'rb') as f: | |
| data = pickle.load(f) | |
| meta = data['meta'].reset_index(drop=True) | |
| image_embedding = data['embedding'] | |
| print(meta.shape, image_embedding.shape) | |
| validation_subset_index = meta['source'].values == 'Val_Tweets' | |
| return meta, image_embedding, validation_subset_index | |
| def app(): | |
| st.title('Text to Image Retrieval') | |
| st.markdown('#### A pathology image search engine that correlate texts directly with images.') | |
| st.caption('Note: The searching query matches images only. The twitter text does not used for searching.') | |
| meta, image_embedding, validation_subset_index = init() | |
| model, processor = load_path_clip() | |
| data_options = ["All twitter data (2006-03-21 — 2023-01-15)", | |
| "Twitter validation data (2022-11-16 — 2023-01-15)"] | |
| st.radio( | |
| "Choose dataset for image retrieval 👉", | |
| key="datapool", | |
| options=data_options, | |
| ) | |
| col1, col2 = st.columns(2) | |
| #query = st.text_input('Search Query', '') | |
| col1_submit = False | |
| show = False | |
| with col1: | |
| # Create selectbox | |
| examples = ['Breast tumor surrounded by fat', | |
| 'HER2+ breast tumor', | |
| 'Colorectal cancer tumor on epithelium', | |
| 'An image of endometrium epithelium', | |
| 'Breast cancer DCIS', | |
| 'Papillary carcinoma in breast tissue', | |
| ] | |
| query_1 = st.selectbox("Please select an example query", options=examples) | |
| #st.info(f":white_check_mark: The written option is {query_1} ") | |
| col1_submit = True | |
| show = True | |
| with col2: | |
| form = st.form(key='my_form') | |
| query_2 = form.text_input(label='Or input your custom query:') | |
| submit_button = form.form_submit_button(label='Submit') | |
| if submit_button: | |
| col1_submit = False | |
| show = True | |
| if col1_submit: | |
| query = query_1 | |
| else: | |
| query = query_2 | |
| text_embedding = embed_texts(model, [query], processor)[0].detach().cpu().numpy() | |
| text_embedding = text_embedding/np.linalg.norm(text_embedding) | |
| similarity_scores = text_embedding.dot(image_embedding.T) | |
| topn = 5 | |
| if st.session_state.datapool == data_options[0]: | |
| #Use all twitter data | |
| id_sorted = np.argsort(similarity_scores)[::-1] | |
| best_ids = id_sorted[:topn] | |
| best_scores = similarity_scores[best_ids] | |
| target_weblinks = meta["weblink"].values[best_ids] | |
| else: | |
| #Use validation twitter data | |
| similarity_scores = similarity_scores[validation_subset_index] | |
| # Sort IDs by cosine-similarity from high to low | |
| id_sorted = np.argsort(similarity_scores)[::-1] | |
| best_ids = id_sorted[:topn] | |
| best_scores = similarity_scores[best_ids] | |
| target_weblinks = meta["weblink"].values[validation_subset_index][best_ids] | |
| #TODO: Avoid duplicated ID | |
| topk_options = ['1st', '2nd', '3rd', '4th', '5th'] | |
| st.radio( | |
| "Choose the most similar 👉", | |
| key="top_k", | |
| options=topk_options, | |
| horizontal=True | |
| ) | |
| topn_txt = st.session_state.top_k | |
| topn_value = int(st.session_state.top_k[0])-1 | |
| st.caption(f'The {topn_txt} relevant image (similarity = {best_scores[topn_value]:.4f})') | |
| components.html(''' | |
| <blockquote class="twitter-tweet"> | |
| <a href="%s"></a> | |
| </blockquote> | |
| <script async src="https://platform.twitter.com/widgets.js" charset="utf-8"> | |
| </script> | |
| ''' % target_weblinks[topn_value], | |
| height=800) | |
| st.markdown('Disclaimer') | |
| st.caption('Please be advised that this function has been developed in compliance with the Twitter policy of data usage and sharing. It is important to note that the results obtained from this function are not intended to constitute medical advice or replace consultation with a qualified medical professional. The use of this function is solely at your own risk and should be consistent with applicable laws, regulations, and ethical considerations. We do not warrant or guarantee the accuracy, completeness, suitability, or usefulness of this function for any particular purpose, and we hereby disclaim any liability arising from any reliance placed on this function or any results obtained from its use. If you wish to review the original Twitter post, you should access the source page directly on Twitter.') | |