import streamlit as st

# Setting the states
def initialize_states():
    # Streamlit state variables
    if "model_name" not in st.session_state:
        st.session_state.model_name = None
    
    if "layer_name" not in st.session_state:
        st.session_state.layer_name = None
    
    if "layer_list" not in st.session_state:
        st.session_state.layer_list = None
    
    if "model" not in st.session_state:
        st.session_state.model = None
    
    if "feat_extract" not in st.session_state:
        st.session_state.feat_extract = None


# Strings
replicate = ":bulb: Choose **ResNet50V2** model and **conv3_block4_out** to get the results as in the example."
credits = ":memo: [Keras example](https://keras.io/examples/vision/visualizing_what_convnets_learn/) by [@fchollet](https://twitter.com/fchollet)."
vit_info = ":star: For Vision Transformers, check the excellent [probing-vits](https://huggingface.co/probing-vits) space."

title = "Visualizing What Convnets Learn"
info_text = """
                Models in this demo are pre-trained on the ImageNet dataset.
                The simple visualization process involves creation of input images that maximize the activation of specific filters in a target layer.
                Such images represent a visualization of the pattern that the filter responds to. 
            """
self_credit = "Space by Vrinda Prabhu"


# Constants and globals
IMG_WIDTH = 180
IMG_HEIGHT = 180
VIS_OPTION = {"only the first filter": 0, "the first 64 filters": 64}
ITERATIONS = 30
LEARNING_RATE = 10.0