app.py

import streamlit as st
from helper import (
    load_dataset, search, get_file_paths,
    get_cordinates, get_images_from_s3_to_display,
    get_images_with_bounding_boxes_from_s3, load_dataset_with_limit
)
import os
import time
import psutil
from memory_profiler import memory_usage

# Load environment variables
AWS_ACCESS_KEY_ID = os.getenv("AWS_ACCESS_KEY_ID")
AWS_SECRET_ACCESS_KEY = os.getenv("AWS_SECRET_ACCESS_KEY")

# Predefined list of datasets
datasets = ["WayveScenes", "MajorTom-Germany"]
description = {
    "WayveScenes": "A large-scale dataset featuring diverse urban driving scenes, captured from autonomous vehicles to advance AI perception and navigation in complex environments.",
    "MajorTom-Germany": "A geospatial dataset containing satellite imagery from across Germany, designed for tasks like land-use classification, environmental monitoring, and earth observation analytics."
}
selection = {
    'WayveScenes': [1, 8],
    "MajorTom-Germany": [1, 1]
}
folder_path_dict = {
        "WayveScenes" : 'WayveScenes/',
        "MajorTom-Germany": "MajorTom-Europe/"
    }
# AWS S3 bucket name
bucket_name = "datasets-quasara-io"

# Function to display CPU and memory usage
def display_usage():
    process = psutil.Process(os.getpid())
    st.write(f"CPU usage: {process.cpu_percent()}%")
    st.write(f"Memory usage: {process.memory_info().rss / (1024 ** 2)} MB")

# Streamlit App
def main():
    # Initialize session state variables if not already initialized
    if 'search_in_small_objects' not in st.session_state:
        st.session_state.search_in_small_objects = False

    if 'dataset_number' not in st.session_state:
        st.session_state.dataset_number = 1

    if 'df' not in st.session_state:
        st.session_state.df = None

    st.title("Semantic Search and Image Display")

    # Select dataset from dropdown
    dataset_name = st.selectbox("Select Dataset", datasets)

    folder_path = folder_path_dict[dataset_name]

    st.caption(description[dataset_name])

    if st.checkbox("Enable Small Object Search", value=st.session_state.search_in_small_objects):
        st.session_state.search_in_small_objects = True
        st.text("Small Object Search Enabled")
        st.session_state.dataset_number = st.selectbox("Select Subset of Data", list(range(1, selection[dataset_name][1] + 1)))
        st.text(f"You have selected Split Dataset {st.session_state.dataset_number}")
    else:
        st.session_state.search_in_small_objects = False
        st.text("Small Object Search Disabled")
        st.session_state.dataset_number = st.selectbox("Select Subset of Data", list(range(1, selection[dataset_name][0] + 1)))
        st.text(f"You have selected Main Dataset {st.session_state.dataset_number}")

    dataset_limit = st.slider("Size of Dataset to be searched from", min_value=1000, max_value=30000, value=10000)
    st.text(f'The smaller the dataset the faster the search will work.')

    # Load dataset with limit only if not already loaded
    if st.button("Load Dataset"):
        try:
            loading_dataset_text = st.empty()
            loading_dataset_text.text("Loading Dataset...")
            loading_dataset_bar = st.progress(0)
            
            # Memory profiling
            mem_usage = memory_usage((load_dataset_with_limit, (dataset_name, st.session_state.dataset_number, st.session_state.search_in_small_objects), {"limit": dataset_limit}))
            st.write(f"Memory used for loading the dataset: {mem_usage[-1]:.2f} MB")
            
            # Simulate dataset loading progress
            for i in range(0, 100, 25):
                time.sleep(0.2)  # Simulate work being done
                loading_dataset_bar.progress(i + 25)

            # Load dataset and monitor CPU and memory
            df, total_rows = load_dataset_with_limit(dataset_name, st.session_state.dataset_number, st.session_state.search_in_small_objects, limit=dataset_limit)
            
            # Store loaded dataset in session state
            st.session_state.df = df
            loading_dataset_bar.progress(100)
            loading_dataset_text.text("Dataset loaded successfully!")
            st.success(f"Dataset loaded successfully with {len(df)} rows.")
            
            # Display CPU and memory usage
            display_usage()
            
        except Exception as e:
            st.error(f"Failed to load dataset: {e}")
    
    
    # Input search query
    query = st.text_input("Enter your search query")

    # Number of results to display
    limit = st.number_input("Number of results to display", min_value=1, max_value=10, value=10)

    # Search button
    if st.button("Search"):
        # Validate input
        if not query:
            st.warning("Please enter a search query.")
        else:
            try:
                # Progress bar for search
                search_loading_text = st.empty()
                search_loading_text.text("Searching...")
                search_progress_bar = st.progress(0)

                # Perform search on the loaded dataset from session state
                df = st.session_state.df
                if st.session_state.search_in_small_objects:
                    results = search(query, df, limit)
                    top_k_paths = get_file_paths(df, results)
                    top_k_cordinates = get_cordinates(df, results)
                else:
                    # Normal Search
                    results = search(query, df, limit)
                    top_k_paths = get_file_paths(df, results)

                # Complete the search progress
                search_progress_bar.progress(100)
                search_loading_text.text("Search completed!")

                # Load Images with Bounding Boxes if applicable
                if st.session_state.search_in_small_objects and top_k_paths and top_k_cordinates:
                    get_images_with_bounding_boxes_from_s3(bucket_name, top_k_paths, top_k_cordinates, AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, folder_path)
                elif not st.session_state.search_in_small_objects and top_k_paths:
                    st.write(f"Displaying top {len(top_k_paths)} results for query '{query}':")
                    get_images_from_s3_to_display(bucket_name, top_k_paths, AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, folder_path)
                    
                else:
                    st.write("No results found.")

                # Display CPU and memory usage
                display_usage()

            except Exception as e:
                st.error(f"Search failed: {e}")

if __name__ == "__main__":
    main()