twitter iframe

.gitattributes

@@ -32,11 +32,3 @@ 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
-*.csv filter=lfs diff=lfs merge=lfs -text
-*.asset filter=lfs diff=lfs merge=lfs -text
-twitter.asset filter=lfs diff=lfs merge=lfs -text
-*.psd filter=lfs diff=lfs merge=lfs -text
-*.jpg filter=lfs diff=lfs merge=lfs -text
-*.jpeg filter=lfs diff=lfs merge=lfs -text
-*.png filter=lfs diff=lfs merge=lfs -text
-*.svg filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -1,2 +0,0 @@
-/__pycache__
-*.pyc

app.py

@@ -1,27 +1,30 @@
+import streamlit as st
+import pandas as pd
 import home
+import numpy as np
+from PIL import Image
+import requests
+import transformers
 import text2image
-import image2image
-import visualization
+import tokenizers
+from io import BytesIO
 import streamlit as st
-import details
-
-st.set_page_config(layout="wide")
+from transformers import CLIPModel
+import clip
+import torch
+from transformers import (
+    VisionTextDualEncoderModel,
+    AutoFeatureExtractor,
+    AutoTokenizer
+)
+from transformers import AutoProcessor
 
-st.sidebar.title("WebPLIP")
-st.sidebar.markdown("## Menu")
+st.sidebar.title("Explore our PLIP Demo")
 
 PAGES = {
     "Introduction": home,
-    "Details": details,
     "Text to Image": text2image,
-    "Image to Image": image2image,
-    "Visualization": visualization,
 }
 
 page = st.sidebar.radio("", list(PAGES.keys()))
-st.sidebar.markdown("## Links")
-
-st.sidebar.markdown("[PLIP Model](https://huggingface.co/vinid/plip)")
-st.sidebar.markdown("[OpenPath Dataset](https://drive.google.com/drive/folders/1b5UT8BzUphkHZavRG-fmiyY9JWYIWZER)")
-st.sidebar.markdown("[PLIP Code](https://github.com/vinid/path_eval)")
 PAGES[page].app()

data/.gitignore

@@ -1 +0,0 @@
-/data_ignore

data/data_weblinks.csv

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:962bb004ef3d4f3ba02659ece1c9d4ae8f61083410345f99ea525d1249662a4d
-size 16380376

data/img_2d_embedding.csv

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:2c94660a66598254f87494fbb931f01a78feda5452d54c3b3939543a392d2fb7
-size 13600550

data/twitter.asset

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:533c5c023e099a2725be0241ca57d6218e37f1f355963ced8d2305270312e428
-size 245669888

data/txt_2d_embedding.csv

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:d5e5cde958eda7b8e74f327326029df1f7480c3bb2879141bdab8f71dd71fdd4
-size 13659850

details.py

@@ -1,73 +0,0 @@
-from pathlib import Path
-import streamlit as st
-import streamlit.components.v1 as components
-from PIL import Image
-import base64
-
-def read_markdown_file(markdown_file):
-    return Path(markdown_file).read_text()
-
-def render_svg(svg_filename):
-    with open(svg_filename,"r") as f:
-        lines = f.readlines()
-        svg=''.join(lines)
-    """Renders the given svg string."""
-    b64 = base64.b64encode(svg.encode('utf-8')).decode("utf-8")
-    html = r'<img src="data:image/svg+xml;base64,%s"/>' % b64
-    st.write(html, unsafe_allow_html=True)
-
-
-def app():
-    #intro_markdown = read_markdown_file("introduction.md")
-    #st.markdown(intro_markdown, unsafe_allow_html=True)
-    st.markdown("# Leveraging medical Twitter to build a visual-language foundation model for pathology")
-
-
-    st.markdown("The lack of annotated publicly available medical images is a major barrier for innovations. At the same time, many de-identified images and much knowledge are shared by clinicians on public forums such as medical Twitter. Here we harness these crowd platforms to curate OpenPath, a large dataset of <b>208,414</b> pathology images paired with natural language descriptions. This is the largest public dataset for pathology images annotated with natural text. We demonstrate the value of this resource by developing PLIP, a multimodal AI with both image and text understanding, which is trained on OpenPath. PLIP achieves state-of-the-art zero-shot and few-short performance for classifying new pathology images across diverse tasks. Moreover, PLIP enables users to retrieve similar cases by either image or natural language search, greatly facilitating knowledge sharing. Our approach demonstrates that publicly shared medical data is a tremendous opportunity that can be harnessed to advance biomedical AI.", unsafe_allow_html=True)
-    render_svg("resources/SVG/Asset 49.svg")
-
-    st.markdown('#### Watch our successful image-to-image retrieval via PLIP:')
-    col1, col2, col3, _, _ = st.columns([1, 1, 1, 1, 1])
-    with col1:
-        st.markdown("[Similar cells](https://twitter.com/ZhiHuangPhD/status/1641906064823312384)")
-        example1 = Image.open('resources/example/1.png')
-        st.image(example1, caption='Example 1', output_format='png')
-    with col2:
-        st.markdown("[Salient object](https://twitter.com/ZhiHuangPhD/status/1641899092195565569)")
-        example2 = Image.open('resources/example/2.png')
-        st.image(example2, caption='Example 2', output_format='png')
-    with col3:
-        st.markdown("[Similar region](https://twitter.com/ZhiHuangPhD/status/1641911235288645632)")
-        example3 = Image.open('resources/example/3.png')
-        st.image(example3, caption='Example 3', output_format='png')
-
-
-    st.markdown("#### PLIP is trained on the largest public vision–language pathology dataset: OpenPath")
-
-    col1, col2 = st.columns([1, 1])
-    with col1:
-        st.markdown("Following the usage policy and guidelines from Twitter and other entities, we established so far the largest public vision–language pathology dataset. To ensure the quality of the data, OpenPath followed rigorous protocols for cohort inclusion and exclusion, including the removal of retweets, sensitive tweets, and non-pathology images, as well as text cleaning.", unsafe_allow_html=True)
-        st.markdown("The final OpenPath dataset consists of:", unsafe_allow_html=True)
-        st.markdown("- Tweets: 116,504 image–text pairs from Twitter posts (tweets) during Mar. 21, 2006 – Nov. 15, 2022 across 32 pathology subspecialty-specific hashtags;", unsafe_allow_html=True)
-        st.markdown("- Replies: 59,869 image–text pairs from the associated replies that received the highest number of likes in the tweet, if applicable;", unsafe_allow_html=True)
-        st.markdown("- PathLAION: 32,041 additional image–text pairs from the Internet which are outside from the Twitter community extracted from the LAION dataset.", unsafe_allow_html=True)
-        st.markdown("Leveraging the largest publicly available pathology dataset which contains image–text pairs across 32 different pathology subspecialty-specific hashtags, where each image has detailed text descriptions, we fine-tuned a pre-trained CLIP model and proposed a multimodal deep learning model for pathology, PLIP.", unsafe_allow_html=True)
-    with col2:
-        render_svg("resources/SVG/Asset 50.svg")
-        render_svg("resources/SVG/Asset 51.svg")
-
-
-
-    st.markdown("#### PLIP is trained with connecting the image and text via contrastive learning")
-
-    col1, col2 = st.columns([3, 1])
-    with col1:
-        st.markdown("The proposed PLIP model generates two embedding vectors from both the text and image encoders. These vectors were then forced to be similar for each of the paired image and text vectors and dissimilar for non-paired image and text pairs via contrastive learning.", unsafe_allow_html=True)
-        fig1e = Image.open('resources/4x/Fig1e.png')
-        st.image(fig1e, caption='PLIP training', output_format='png')
-
-    with col2:
-        render_svg("resources/SVG/Asset 53.svg")
-
-
-

home.py

@@ -1,52 +1,11 @@
 from pathlib import Path
 import streamlit as st
-import streamlit.components.v1 as components
-from PIL import Image
-import base64
+
 
 def read_markdown_file(markdown_file):
     return Path(markdown_file).read_text()
 
-def render_svg(svg_filename):
-    with open(svg_filename,"r") as f:
-        lines = f.readlines()
-        svg=''.join(lines)
-    """Renders the given svg string."""
-    b64 = base64.b64encode(svg.encode('utf-8')).decode("utf-8")
-    html = r'<img src="data:image/svg+xml;base64,%s"/>' % b64
-    st.write(html, unsafe_allow_html=True)
-
 
 def app():
-
-    st.markdown("# A visual-language foundation model for pathology")
-    st.markdown("This is a webapp for PLIP, our new fundational AI model for pathology and OpenPath our new dataset, **from our recent work**: Leveraging medical Twitter to build a visual-language foundation model for pathology")
-    st.markdown("### Pathology Language and Image Pretraining (PLIP)\n We develop PLIP, a multimodal AI with both image and text understanding. PLIP achieves state-of-the-art zero-shot and few-short performance for classifying new pathology images across diverse tasks. Moreover, PLIP enables users to retrieve similar cases by either image or natural language search, greatly facilitating knowledge sharing. Our approach demonstrates that publicly shared medical data is a tremendous opportunity that can be harnessed to advance biomedical AI.")
-
-    fig1e = Image.open('resources/4x/Fig1e.png')
-    st.image(fig1e, caption='PLIP training procedure', output_format='png')
-
-    st.markdown("### OpenPath Dataset\nThe lack of annotated publicly available medical images is a major barrier for innovations. At the same time, many de-identified images and much knowledge are shared by clinicians on public forums such as medical Twitter. Here we harness these crowd platforms to curate OpenPath, a large dataset of **208,414** pathology images paired with natural language descriptions")
-
-    render_svg("resources/SVG/Asset 49.svg")
-
-
-    st.markdown("### Documentation\n"
-                "This webapp comes with different functionalities.\n"
-                "* Details: The details page guides you through our work.\n"
-                "* Text to Image: allows users to perform text search on a database of images.\n"
-                "* Image to Image: allows users to perform image search on a database of images.\n"
-                "")
-
-    st.markdown("### Other Links\n"
-                "* Download [OpenPath](https://drive.google.com/drive/folders/1b5UT8BzUphkHZavRG-fmiyY9JWYIWZER)\n"
-                "* Code to reproduce [PLIP](https://github.com/vinid/path_eval) results\n"
-                "* Link to the [PLIP Model](https://huggingface.co/vinid/plip)\n"
-                "")
-
-    st.markdown("""---""")
-    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.')
-
-    st.markdown('Privacy statement')
-    st.caption('In accordance with the privacy and control policy of Twitter, we hereby declared that the data redistributed by us shall only comprise of Tweet IDs. The Tweet IDs will be employed to establish a linkage with the original Twitter post, as long as the original post is still accessible. The hyperlink will cease to function if the user deletes the original post. It is important to note that all tweets displayed on our service have already been classified as non-sensitive by Twitter. It is strictly prohibited to redistribute any content apart from the Tweet IDs. Any distribution carried out must adhere to the laws and regulations applicable in your jurisdiction, including export control laws and embargoes.')
+    intro_markdown = read_markdown_file("introduction.md")
+    st.markdown(intro_markdown, unsafe_allow_html=True)

image2image.py

@@ -1,239 +0,0 @@
-import streamlit as st
-import pandas as pd
-import numpy as np
-from PIL import Image
-import requests
-import tokenizers
-import os
-from io import BytesIO
-import pickle
-import base64
-import datetime
-
-import torch
-from transformers import (
-    VisionTextDualEncoderModel,
-    AutoFeatureExtractor,
-    AutoTokenizer,
-    CLIPModel,
-    AutoProcessor
-)
-import streamlit.components.v1 as components
-from st_clickable_images import clickable_images #pip install st-clickable-images
-
-
-@st.cache(
-    hash_funcs={
-        torch.nn.parameter.Parameter: lambda _: None,
-        tokenizers.Tokenizer: lambda _: None,
-        tokenizers.AddedToken: lambda _: None
-    }
-)
-def load_path_clip():
-    model = CLIPModel.from_pretrained("vinid/plip")
-    processor = AutoProcessor.from_pretrained("vinid/plip")
-    return model, processor
-
-@st.cache
-def init():
-    with open('data/twitter.asset', 'rb') as f:
-        data = pickle.load(f)
-    meta = data['meta'].reset_index(drop=True)
-    image_embedding = data['image_embedding']
-    text_embedding = data['text_embedding']
-    print(meta.shape, image_embedding.shape)
-    validation_subset_index = meta['source'].values == 'Val_Tweets'
-    return meta, image_embedding, text_embedding, validation_subset_index
-
-def embed_images(model, images, processor):
-    inputs = processor(images=images)
-    pixel_values = torch.tensor(np.array(inputs["pixel_values"]))
-
-    with torch.no_grad():
-        embeddings = model.get_image_features(pixel_values=pixel_values)
-    return embeddings
-
-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 app():
-    st.title('Image to Image Retrieval')
-    st.markdown('#### A pathology image search engine that correlate images with images.')
-    st.markdown("Image-to-image retrieval can be used to retrieve pathology images that have contents similar to the target image input, with the ability to comprehend the key components from the input image.")
-    
-    st.markdown('#### Demo')
-
-    meta, image_embedding, text_embedding, validation_subset_index = init()
-    model, processor = load_path_clip()
-
-
-    col1, col2 = st.columns(2)
-    with col1:
-        data_options = ["All twitter data (03/21/2006 — 01/15/2023)",
-                        "Twitter validation data (11/16/2022 — 01/15/2023)"]
-        st.radio(
-            "Choose dataset for image retrieval 👉",
-            key="datapool",
-            options=data_options,
-        )
-    with col2:
-        retrieval_options = ["Image only",
-                            "Text and image (beta)",
-                             ]
-        st.radio(
-            "Similarity calcuation 👉",
-            key="calculation_option",
-            options=retrieval_options,
-        )
-
-
-    st.markdown('Try out following examples:')
-    example_path = 'data/example_images'
-    list_of_examples = [os.path.join(example_path, v) for v in os.listdir(example_path)]
-    example_imgs = []
-    for file in list_of_examples:
-        with open(file, "rb") as image:
-            encoded = base64.b64encode(image.read()).decode()
-            example_imgs.append(f"data:image/jpeg;base64,{encoded}")
-    clicked = clickable_images(
-        example_imgs,
-        titles=[f"Image #{str(i)}" for i in range(len(example_imgs))],
-        div_style={"display": "flex", "justify-content": "center", "flex-wrap": "wrap"},
-        img_style={"margin": "5px", "height": "70px"},
-    )
-    isExampleClicked = False
-    if clicked > -1:
-        image = Image.open(list_of_examples[clicked])
-        isExampleClicked = True
-        
-
-
-
-
-
-    col1, col2, _ = st.columns(3)
-    with col1:
-        query = st.file_uploader("Choose a file to upload")
-
-
-    proceed = False
-    if query:
-        image = Image.open(query)
-        proceed = True
-    elif isExampleClicked:
-        proceed = True
-
-    if proceed:
-        with col2:
-            st.image(image, caption='Your upload')
-
-        input_image = embed_images(model, [image], processor)[0].detach().cpu().numpy()
-
-        input_image = input_image/np.linalg.norm(input_image)
-        
-        # Sort IDs by cosine-similarity from high to low
-
-        if st.session_state.calculation_option == retrieval_options[0]: # Image only
-            similarity_scores = input_image.dot(image_embedding.T)
-        else: # Text and Image
-            similarity_scores_i = input_image.dot(image_embedding.T)
-            similarity_scores_t = input_image.dot(text_embedding.T)
-            similarity_scores_i = similarity_scores_i/np.max(similarity_scores_i)
-            similarity_scores_t = similarity_scores_t/np.max(similarity_scores_t)
-            similarity_scores = (similarity_scores_i + similarity_scores_t)/2
-
-
-        ############################################################
-        # Get top results
-        ############################################################
-        topn = 5
-        df = pd.DataFrame(np.c_[np.arange(len(meta)), similarity_scores, meta['weblink'].values], columns = ['idx', 'score', 'twitterlink'])
-        if st.session_state.datapool == data_options[1]: #Use val twitter data
-            df = df.loc[validation_subset_index,:]
-        df = df.sort_values('score', ascending=False)
-        df = df.drop_duplicates(subset=['twitterlink'])
-        best_id_topk = df['idx'].values[:topn]
-        target_scores = df['score'].values[:topn]
-        target_weblinks = df['twitterlink'].values[:topn]
-
-
-
-        ############################################################
-        # Display results
-        ############################################################
-        
-        st.markdown('#### Top 5 results:')
-        topk_options = ['1st', '2nd', '3rd', '4th', '5th']
-        tab = {}
-        tab[0], tab[1], tab[2] = st.columns(3)
-        for i in [0,1,2]:
-            with tab[i]:
-                topn_value = i
-                topn_txt = topk_options[i]
-                st.caption(f'The {topn_txt} relevant image (similarity = {target_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=600)
-
-        tab[3], tab[4], tab[5] = st.columns(3)
-        for i in [3,4]:
-            with tab[i]:
-                topn_value = i
-                topn_txt = topk_options[i]
-                st.caption(f'The {topn_txt} relevant image (similarity = {target_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("""---""")
-    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.')
-
-    st.markdown('Privacy statement')
-    st.caption('In accordance with the privacy and control policy of Twitter, we hereby declared that the data redistributed by us shall only comprise of Tweet IDs. The Tweet IDs will be employed to establish a linkage with the original Twitter post, as long as the original post is still accessible. The hyperlink will cease to function if the user deletes the original post. It is important to note that all tweets displayed on our service have already been classified as non-sensitive by Twitter. It is strictly prohibited to redistribute any content apart from the Tweet IDs. Any distribution carried out must adhere to the laws and regulations applicable in your jurisdiction, including export control laws and embargoes.')
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-

introduction.md

@@ -0,0 +1,2 @@
+
+# Welcome to our PLIP Demo

plip_support.py

@@ -0,0 +1,9 @@
+import clip
+import torch
+
+
+
+
+def embed_text(plip, text, device="cpu"):
+    idx = clip.tokenize([text], truncate=True).to(device)
+    return plip.encode_text(idx).detach().cpu().numpy()[0]

requirements.txt

@@ -1,8 +1,7 @@
-
-tokenizers
-pandas
+git+https://github.com/openai/CLIP.git
 torch
 transformers
-st_clickable_images
-plotly
-altair<5.0
+pandas
+numpy
+Pillow
+streamlit==1.19.0

text2image.py

@@ -1,58 +1,23 @@
 import streamlit as st
 import pandas as pd
+from plip_support import embed_text
 import numpy as np
 from PIL import Image
-import pickle
+import requests
+import transformers
 import tokenizers
+from io import BytesIO
+import streamlit as st
+from transformers import CLIPModel
+import clip
 import torch
-
 from transformers import (
-
-    CLIPModel,
-    AutoProcessor
+    VisionTextDualEncoderModel,
+    AutoFeatureExtractor,
+    AutoTokenizer
 )
-import streamlit.components.v1 as components
-import base64
-
-def render_svg(svg_filename):
-    with open(svg_filename,"r") as f:
-        lines = f.readlines()
-        svg=''.join(lines)
-    """Renders the given svg string."""
-    b64 = base64.b64encode(svg.encode('utf-8')).decode("utf-8")
-    html = r'<img src="data:image/svg+xml;base64,%s"/>' % b64
-    st.write(html, unsafe_allow_html=True)
-
-@st.cache(
-    hash_funcs={
-        torch.nn.parameter.Parameter: lambda _: None,
-        tokenizers.Tokenizer: lambda _: None,
-        tokenizers.AddedToken: lambda _: None
-    }
-)
-def load_path_clip():
-    model = CLIPModel.from_pretrained("vinid/plip")
-    processor = AutoProcessor.from_pretrained("vinid/plip")
-    return model, processor
+from transformers import AutoProcessor
 
-@st.cache
-def init():
-    with open('data/twitter.asset', 'rb') as f:
-        data = pickle.load(f)
-    meta = data['meta'].reset_index(drop=True)
-    image_embedding = data['image_embedding']
-    text_embedding = data['text_embedding']
-    print(meta.shape, image_embedding.shape)
-    validation_subset_index = meta['source'].values == 'Val_Tweets'
-    return meta, image_embedding, text_embedding, validation_subset_index
-
-def embed_images(model, images, processor):
-    inputs = processor(images=images)
-    pixel_values = torch.tensor(np.array(inputs["pixel_values"]))
-
-    with torch.no_grad():
-        embeddings = model.get_image_features(pixel_values=pixel_values)
-    return embeddings
 
 def embed_texts(model, texts, processor):
     inputs = processor(text=texts, padding="longest")
@@ -65,162 +30,45 @@ def embed_texts(model, texts, processor):
         )
     return embeddings
 
+@st.cache
+def load_embeddings(embeddings_path):
+    print("loading embeddings")
+    return np.load(embeddings_path)
 
-def app():
-
-    st.title('Text to Image Retrieval')
-    st.markdown('#### A pathology image search engine that geos from texts to images.')
-    
-    col1, col2 = st.columns([1,1])
-    with col1:
-        st.markdown("The text-to-image retrieval system can serve as an image search engine, enabling users to match images from multiple queries and retrieve the most relevant image based on a sentence description. This generic system can comprehend semantic and interrelated knowledge, such as “Breast tumor surrounded by fat”.")
-        st.markdown("Unlike searching keywords and sentences from Google and indirectly matching the images from the target text, our proposed pathology image retrieval allows direct comparison between input sentences and images.")
-    with col2:
-        render_svg("resources/SVG/Asset 54.svg")
-
-    meta, image_embedding, text_embedding, validation_subset_index = init()
-    model, processor = load_path_clip()
-
-    st.markdown('### Search')
-    st.markdown('How to use this: first of all, select a dataset on which to do retrieval.\n'
-                'Then, either select a predefined search query or input one yourself.')
-
-
-    col1, col2 = st.columns(2)
-    with col1:
-        data_options = ["All Twitter Data (03/21/2006 — 01/15/2023)",
-                        "Validation Twitter data (11/16/2022 — 01/15/2023)"]
-        st.selectbox(
-            "Dataset",
-            key="datapool",
-            options=data_options,
-        )
-
-    with col2:
-        retrieval_options = ["Image only",
-                             "Text and image (beta)",
-                             ]
-        st.radio(
-            "Similarity calcuation 👉",
-            key="calculation_option",
-            options=retrieval_options,
-        )
-
-    col1, col2 = st.columns(2)
-
-    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("Select an example", options=examples)
-
-        col1_submit = 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
-
-    if col1_submit:
-        query = query_1
-    else:
-        query = query_2
-
-    input_text = embed_texts(model, [query], processor)[0].detach().cpu().numpy()
-    input_text = input_text/np.linalg.norm(input_text)
-
-    # Sort IDs by cosine-similarity from high to low
-
-    if st.session_state.calculation_option == retrieval_options[0]:  # Image only
-        similarity_scores = input_text.dot(image_embedding.T)
-    else:  # Text and Image
-        similarity_scores_i = input_text.dot(image_embedding.T)
-        similarity_scores_t = input_text.dot(text_embedding.T)
-        similarity_scores_i = similarity_scores_i / np.max(similarity_scores_i)
-        similarity_scores_t = similarity_scores_t / np.max(similarity_scores_t)
-        similarity_scores = (similarity_scores_i + similarity_scores_t) / 2
-
-    ############################################################
-    # Get top results
-    ############################################################
-    topn = 5
-    df = pd.DataFrame(np.c_[np.arange(len(meta)), similarity_scores, meta['weblink'].values], columns = ['idx', 'score', 'twitterlink'])
-    if st.session_state.datapool == data_options[1]: #Use val twitter data
-        df = df.loc[validation_subset_index,:]
-    df = df.sort_values('score', ascending=False)
-    df = df.drop_duplicates(subset=['twitterlink'])
-    best_id_topk = df['idx'].values[:topn]
-    target_scores = df['score'].values[:topn]
-    target_weblinks = df['twitterlink'].values[:topn]
-
-
-    ############################################################
-    # Display results
-    ############################################################
-
-    text = '<font size="4">Your input query: <span style="background-color: rgb(230,230,230);"><b>%s</b></span>' % query + \
-        ' (Try search it directly on [Twitter](https://twitter.com/search?q=%s&src=typed_query) or [Google](https://www.google.com/search?q=%s))</font>' % (query.replace(' ', '%20'), query.replace(' ', '+'))
-    st.markdown(text, unsafe_allow_html=True)
-
-    st.markdown('#### Top 5 results:')
-    topk_options = ['1st', '2nd', '3rd', '4th', '5th']
-    tab = {}
-    tab[0], tab[1], tab[2] = st.columns(3)
-    for i in [0,1,2]:
-        with tab[i]:
-            topn_value = i
-            topn_txt = topk_options[i]
-            st.caption(f'The {topn_txt} relevant image (similarity = {target_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=600)
-
-    tab[3], tab[4], tab[5] = st.columns(3)
-    for i in [3,4]:
-        with tab[i]:
-            topn_value = i
-            topn_txt = topk_options[i]
-            st.caption(f'The {topn_txt} relevant image (similarity = {target_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("""---""")
-    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.')
-
-    st.markdown('Privacy statement')
-    st.caption('In accordance with the privacy and control policy of Twitter, we hereby declared that the data redistributed by us shall only comprise of Tweet IDs. The Tweet IDs will be employed to establish a linkage with the original Twitter post, as long as the original post is still accessible. The hyperlink will cease to function if the user deletes the original post. It is important to note that all tweets displayed on our service have already been classified as non-sensitive by Twitter. It is strictly prohibited to redistribute any content apart from the Tweet IDs. Any distribution carried out must adhere to the laws and regulations applicable in your jurisdiction, including export control laws and embargoes.')
-
+@st.cache(
+    hash_funcs={
+        torch.nn.parameter.Parameter: lambda _: None,
+        tokenizers.Tokenizer: lambda _: None,
+        tokenizers.AddedToken: lambda _: None
+    }
+)
+def load_path_clip():
+    model = CLIPModel.from_pretrained("vinid/plip")
+    processor = AutoProcessor.from_pretrained("vinid/plip")
+    return model, processor
 
 
+def app():
+    st.title('PLIP Image Search')
 
+    plip_dataset = pd.read_csv("tweet_eval_retrieval.tsv", sep="\t")
 
+    model, processor = load_path_clip()
 
+    image_embedding = load_embeddings("tweet_eval_embeddings.npy")
 
+    query = st.text_input('Search Query', '')
 
 
+    if query:
 
+        text_embedding = embed_texts(model, [query], processor)[0].detach().cpu().numpy()
 
+        text_embedding = text_embedding/np.linalg.norm(text_embedding)
 
+        best_id = np.argmax(text_embedding.dot(image_embedding.T))
+        url = (plip_dataset.iloc[best_id]["imageURL"])
 
+        response = requests.get(url)
+        img = Image.open(BytesIO(response.content))
+        st.image(img)

data/example_images/Books_0f25e3de-2c23-46df-ac67-b7bcddd5e1e6.png → tweet_eval_embeddings.npy

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bf7d4b43c2bcb233b4869c142c9872efb557888091682d7030a016e47dc0d95c
-size 554136
+oid sha256:36e445b069b1d937a0a780ddeab9239df5fd13264e8cd1f6cf033be3210352e1
+size 2401408