feat: add pages

app.py

@@ -1,16 +1,28 @@
 import streamlit as st
+from st_pages import Page, show_pages, add_page_title, Section 
 from lib.utils.model import get_model, get_similarities 
-from PIL import Image 
 
-st.set_page_config('IRRA Space')
-st.title('IRRA Text-To-Image-Retrival')
-st.markdown('A text-to-image retrieval model implemented from [arXiv: Cross-Modal Implicit Relation Reasoning and Aligning for Text-to-Image Person Retrieval](https://arxiv.org/abs/2303.12501)')
+add_page_title()
+
+show_pages(
+    [
+        Page('app.py', 'IRRA Text-To-Image-Retrival'),
+        Section('Implementation Details'),
+        Page('pages/losses.py', 'Loss functions'),
+    ]
+)
+
+st.markdown('''
+            A text-to-image retrieval model implemented from [arXiv: Cross-Modal Implicit Relation Reasoning and Aligning for Text-to-Image Person Retrieval](https://arxiv.org/abs/2303.12501).
+            The uploaded images should be `384x128` with only one person in the shot. 
+            ''')
 
 st.header('Inputs')
 caption = st.text_input('Description Input')
 
 images = st.file_uploader('Upload images', accept_multiple_files=True)
 if images is not None:
+    
     st.image(images) # type: ignore
 
 st.header('Options')
@@ -32,12 +44,27 @@ if button:
 
     indices = similarities.argsort(descending=True).cpu().tolist()[:ranks]
     
+    c1, c2, c3 = st.columns(3)
+    with c1:
+        st.subheader('Rank')
+    with c2:
+        st.subheader('Image')
+    with c3:
+        st.subheader('Cosine Similarity', help='Due to the nature of the SDM loss, the higher the similarity, the more similar the match is')
+    
     for i, idx in enumerate(indices):
         c1, c2, c3 = st.columns(3)
         with c1:
-            st.text(f'Rank {i + 1}')
+            st.text(f'{i + 1}')
         with c2:
             st.image(images[idx])
         with c3:
-            st.text(f'Cosine sim {similarities[idx].cpu():.2f}')
-    
+            st.text(f'{similarities[idx].cpu():.2f}')
+
+with st.sidebar:
+    st.title('IRRA Text-To-Image Retrival')
+
+    st.subheader('Useful Links')
+    st.markdown('[arXiv: Cross-Modal Implicit Relation Reasoning and Aligning for Text-to-Image Person Retrieval](https://arxiv.org/abs/2303.12501)')
+    st.markdown('[IRRA implementation (Pytorch Lightning + Transformers)](https://github.com/grostaco/modern-IRRA)')
+    st.markdown('[IRRA implementation (PyTorch)](https://github.com/anosorae/IRRA/tree/main)')