| import streamlit as st | |
| from utils.frontend import build_sidebar | |
| build_sidebar() | |
| st.markdown(""" | |
| # Better Image Retrieval With Reinforced CLIP 🧠 | |
| CLIP is a neural network trained on image-text pairs that can predict how semantically close images are with some text | |
| But, although CLIP understands what it sees, it doesn't know its properties. While other models can understand text that contains such information, like Wikipedia. | |
| In this demo application, we see if we can 'help' CLIP by reinforcing it with another model.""") | |
| st.image("diagram.png") |