import streamlit as st import numpy as np import pandas as pd from PIL import Image import cv2 from scipy.ndimage import gaussian_filter import tensorflow as tf import matplotlib.pyplot as plt import io from matplotlib.figure import Figure import base64 # ------------------ TC CENTERING UTILS ------------------ def find_tc_center(ir_image, smoothing_sigma=3): smoothed_image = gaussian_filter(ir_image, sigma=smoothing_sigma) min_coords = np.unravel_index(np.argmin(smoothed_image), smoothed_image.shape) return min_coords[::-1] # Return as (x, y) # Function to generate comparison chart def generate_comparison_chart(models, mae_values, rmse_values, predicted_values=None): # Calculate improvement percentages relative to the first model baseline_mae = mae_values[0] baseline_rmse = rmse_values[0] mae_improvements = [0] + [((baseline_mae - val) / baseline_mae) * 100 for val in mae_values[1:]] rmse_improvements = [0] + [((baseline_rmse - val) / baseline_rmse) * 100 for val in rmse_values[1:]] # Create figure with subplots (2 or 3 depending on if we have predictions) if predicted_values: fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(18, 8)) else: fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(16, 8)) # Plot MAE bars1 = ax1.bar(range(len(models)), mae_values, color='skyblue', edgecolor='black') ax1.set_title('Mean Absolute Error (MAE)', fontsize=14, fontweight='bold') ax1.set_ylabel('MAE (knots)', fontsize=12) ax1.set_xticks(range(len(models))) ax1.set_xticklabels(models, fontsize=12, rotation=45, ha='right') ax1.grid(axis='y', linestyle='--', alpha=0.3, color='lightgray') ax1.set_ylim(0, max(mae_values) * 1.2) # Plot RMSE bars2 = ax2.bar(range(len(models)), rmse_values, color='lightcoral', edgecolor='black') ax2.set_title('Root Mean Square Error (RMSE)', fontsize=14, fontweight='bold') ax2.set_ylabel('RMSE (knots)', fontsize=12) ax2.set_xticks(range(len(models))) ax2.set_xticklabels(models, fontsize=12, rotation=45, ha='right') ax2.grid(axis='y', linestyle='--', alpha=0.3, color='lightgray') ax2.set_ylim(0, max(rmse_values) * 1.2) # Add values on top of the bars for MAE for i, bar in enumerate(bars1): height = bar.get_height() ax1.text(bar.get_x() + bar.get_width()/2., height + 0.3, f'{height:.2f}', ha='center', va='bottom', fontsize=12) # Add improvement percentage for all except the first bar if i > 0: ax1.text(bar.get_x() + bar.get_width()/2., height/2, f'↓{mae_improvements[i]:.1f}%', ha='center', va='center', color='blue', fontsize=12, fontweight='bold') # Add values on top of the bars for RMSE for i, bar in enumerate(bars2): height = bar.get_height() ax2.text(bar.get_x() + bar.get_width()/2., height + 0.3, f'{height:.2f}', ha='center', va='bottom', fontsize=12) # Add improvement percentage for all except the first bar if i > 0: ax2.text(bar.get_x() + bar.get_width()/2., height/2, f'↓{rmse_improvements[i]:.1f}%', ha='center', va='center', color='darkred', fontsize=12, fontweight='bold') # Add horizontal reference lines for best performance min_mae = min(mae_values) min_rmse = min(rmse_values) ax1.axhline(y=min_mae, color='blue', linestyle='--', alpha=0.5) ax2.axhline(y=min_rmse, color='red', linestyle='--', alpha=0.5) # Add predictions comparison if provided if predicted_values: bars3 = ax3.bar(range(len(models)), predicted_values, color='lightgreen', edgecolor='black') ax3.set_title('Predicted Vmax', fontsize=14, fontweight='bold') ax3.set_ylabel('Wind Speed (knots)', fontsize=12) ax3.set_xticks(range(len(models))) ax3.set_xticklabels(models, fontsize=12, rotation=45, ha='right') ax3.grid(axis='y', linestyle='--', alpha=0.3, color='lightgray') # Add values on top of the bars for predictions for i, bar in enumerate(bars3): height = bar.get_height() ax3.text(bar.get_x() + bar.get_width()/2., height + 0.3, f'{height:.2f}', ha='center', va='bottom', fontsize=12) # Add a label at the bottom explaining the reduction percentages fig.text(0.5, 0.01, 'Note: Reduction percentages (↓%) are calculated relative to TCIP-Net (3DCNN)', ha='center', fontsize=12, fontstyle='italic') plt.tight_layout(rect=[0, 0.03, 1, 0.95]) return fig # Function to convert matplotlib figure to Streamlit-compatible image def fig_to_streamlit(fig): buf = io.BytesIO() fig.savefig(buf, format='png', dpi=300, bbox_inches='tight') buf.seek(0) return buf def extract_local_region(ir_image, center, region_size=95): h, w = ir_image.shape half_size = region_size // 2 x_min = max(center[0] - half_size, 0) x_max = min(center[0] + half_size, w) y_min = max(center[1] - half_size, 0) y_max = min(center[1] + half_size, h) region = np.full((region_size, region_size), np.nan) extracted = ir_image[y_min:y_max, x_min:x_max] region[:extracted.shape[0], :extracted.shape[1]] = extracted return region def generate_hovmoller(X_data): hovmoller_list = [] for ir_images in X_data: # ir_images: shape (8, 95, 95) time_steps = ir_images.shape[0] hovmoller_data = np.zeros((time_steps, 95, 95)) for t in range(time_steps): tc_center = find_tc_center(ir_images[t]) hovmoller_data[t] = extract_local_region(ir_images[t], tc_center, 95) hovmoller_list.append(hovmoller_data) return np.array(hovmoller_list) def reshape_vmax(vmax_values, chunk_size=8): trimmed_size = (len(vmax_values) // chunk_size) * chunk_size vmax_values_trimmed = vmax_values[:trimmed_size] return vmax_values_trimmed.reshape(-1, chunk_size) def create_3d_vmax(vmax_2d_array): # Initialize a 3D array of shape (N, 8, 8) filled with zeros vmax_3d_array = np.zeros((vmax_2d_array.shape[0], 8, 8)) # Fill the diagonal for each row in the 3D array for i in range(vmax_2d_array.shape[0]): np.fill_diagonal(vmax_3d_array[i], vmax_2d_array[i]) # Reshape to (N*8, 8, 8, 1) vmax_3d_array = vmax_3d_array.reshape(-1, 8, 8, 1) # Trim last element if needed (original comment, but not implemented) return vmax_3d_array def process_lat_values(data): lat_values = data # Convert to NumPy array # Trim the array to make its length divisible by 8 trimmed_size = (len(lat_values) // 8) * 8 lat_values_trimmed = lat_values[:trimmed_size] lat_values_trimmed=np.array(lat_values_trimmed) # Convert to NumPy array # Reshape into a 2D array (rows of 8 values each) and remove the last row lat_2d_array = lat_values_trimmed.reshape(-1, 8) return lat_2d_array def process_lon_values(data): lon_values =data # Convert to NumPy array lon_values = np.array(lon_values) # Convert to NumPy array # Trim the array to make its length divisible by 8 trimmed_size = (len(lon_values) // 8) * 8 lon_values_trimmed = lon_values[:trimmed_size] # Reshape into a 2D array (rows of 8 values each) and remove the last row lon_2d_array = lon_values_trimmed.reshape(-1, 8) return lon_2d_array import numpy as np def calculate_intensity_difference(vmax_2d_array): """Calculates intensity difference for each row in Vmax 2D array.""" int_diff = [] for i in vmax_2d_array: k = abs(i[0] - i[-1]) # Absolute difference between first & last element i = np.append(i, k) # Append difference as the 9th element int_diff.append(i) return np.array(int_diff) import numpy as np # Function to process and reshape image data def process_images(images, batch_size=8, img_size=(95, 95, 1)): num_images = images.shape[0] # Trim the dataset to make it divisible by batch_size trimmed_size = (num_images // batch_size) * batch_size images_trimmed = images[:trimmed_size] # Reshape into (x, batch_size, img_size[0], img_size[1], img_size[2]) images_reshaped = images_trimmed.reshape(-1, batch_size, *img_size) return images_reshaped import numpy as np def process_cc_mask(cc_data): """Processes CC mask images by trimming and reshaping into (x, 8, 95, 95, 1).""" num_images = cc_data.shape[0] batch_size = 8 trimmed_size = (num_images // batch_size) * batch_size # Ensure divisibility by 8 images_trimmed = cc_data[:trimmed_size] # Trim excess images cc_images = images_trimmed.reshape(-1, batch_size, 95, 95, 1) # Reshape return cc_images def extract_convective_cores(ir_data): """ Extract Convective Cores (CCs) from IR imagery based on the criteria in the paper. Args: ir_data: IR imagery of shape (height, width). Returns: cc_mask: Binary mask of CCs (1 for CC, 0 otherwise) of shape (height, width). """ height, width,c = ir_data.shape cc_mask = np.zeros_like(ir_data, dtype=np.float32) # Initialize CC mask # Define the neighborhood (8-connected) neighbors = [(-1, -1), (-1, 0), (-1, 1), (0, -1), (0,0) , (0, 1), (1, -1), (1, 0), (1, 1)] for i in range(1, height - 1): # Avoid boundary pixels for j in range(1, width - 1): bt_ij = ir_data[i, j] # Condition 1: BT < 253K if (bt_ij >= 253).any(): continue # Condition 2: BT <= BT_n for all neighbors is_local_min = True for di, dj in neighbors: if ir_data[i + di, j + dj] < bt_ij: is_local_min = False break if not is_local_min: continue # Condition 3: Gradient condition numerator1 = (ir_data[i - 1, j] + ir_data[i + 1, j] - 2 * bt_ij) / 3.1 numerator2 = (ir_data[i, j - 1] + ir_data[i, j + 1] - 2 * bt_ij) / 8.0 lhs = numerator1 + numerator2 rhs = (4 / 5.8) * np.exp(0.0826 * (bt_ij - 217)) if lhs > rhs: cc_mask[i, j] = 1 # Mark as CC return cc_mask def compute_convective_core_masks(ir_data): """Extracts convective core masks for each IR image.""" cc_mask = [] for i in ir_data: c = extract_convective_cores(i) # Assuming this function is defined c = np.array(c) cc_mask.append(c) return np.array(cc_mask) # ------------------ Streamlit UI ------------------ # Configure the page with wide layout and custom title st.set_page_config( page_title="Tropical Cyclone U-Net Wind Speed Predictor", layout="wide", initial_sidebar_state="expanded" ) # Main title with emoji and styling st.markdown("

🌀 Tropical Cyclone U-Net Wind Speed (Intensity) Predictor


", unsafe_allow_html=True) # Authors section with ORCID links st.markdown("""

By: Dharun Krishna K B, Nanduri Prudhvi Reddy and Ravipati Venkata Madan Mohan; School of Computing.
Under the guidance of: Dr. Gowri L, Assistant Professor, School of Computing.
SASTRA Deemed University, Thanjavur, Tamil Nadu, India.

For: Main project titled "Tropical Cyclone Intensity Prediction Using Deep Learning Models"
May 2025

""", unsafe_allow_html=True) # Add a divider before the main content st.markdown('
', unsafe_allow_html=True) # Add spacing st.markdown("
", unsafe_allow_html=True) # App description st.info('''The *Tropical Cyclone Wind Speed Predictor interface* enables the prediction of maximum sustained wind speeds of tropical cyclones (in knots) using IR and PMW imagery, along with physical attributes from the past 24 hours, while also facilitating comparison between state-of-the-art models and our proposed model. ''') ir_images = st.file_uploader("Upload 8 IR images", type=["jpg", "jpeg", "png"], accept_multiple_files=True) pmw_images = st.file_uploader("Upload 8 PMW images", type=["jpg", "jpeg", "png"], accept_multiple_files=True) if len(ir_images) != 8 or len(pmw_images) != 8: st.warning("Please upload exactly 8 IR and 8 PMW images.") else: st.success("Uploaded 8 IR and 8 PMW images successfully.") st.header("Input Latitude, Longitude, Vmax") lat_values, lon_values, vmax_values = [], [], [] import pandas as pd import numpy as np # File uploader csv_file = st.file_uploader("Upload CSV file", type=["csv"]) if csv_file is not None: try: df = pd.read_csv(csv_file) required_columns = {'Latitude', 'Longitude', 'Vmax'} if required_columns.issubset(df.columns): lat_values = df['Latitude'].values lon_values = df['Longitude'].values vmax_values = df['Vmax'].values lat_values = np.array(lat_values) lon_values = np.array(lon_values) vmax_values = np.array(vmax_values) st.success("CSV file loaded and processed successfully!") # Display the dataframe in a scrollable container st.markdown("

Preview of uploaded data:

", unsafe_allow_html=True) preview_df = df.head(10).reset_index(drop=True) preview_df.index += 1 # Shift index to start from 1 st.dataframe(preview_df, height=200) else: st.error("CSV file must contain 'Latitude', 'Longitude', and 'Vmax' columns.") except Exception as e: st.error(f"Error reading CSV: {e}") else: st.warning("Please upload a CSV file.") # Define data for ablation study ablation_data = { "Model": [ "TCIP-Net (3DCNN)", "TCIP-Net (ST-LSTM)", "TCIP-Net (ConvLSTM)", "TCIP-Net (TrajGRU)", "TCIP-Net (ConvGRU)", "TCUWSP-Net (Proposed)" ], "RMSE": [12.63, 12.52, 12.36, 12.24, 11.17, 8.6549], "MAE": [10.15, 10.12, 9.97, 9.93, 8.92, 6.309] } # Improved Prediction Model Section with better UI st.markdown("
", unsafe_allow_html=True) st.markdown("

Select Prediction Model

", unsafe_allow_html=True) # Create columns for better layout col1, col2, col3 = st.columns([1, 2, 1]) with col2: model_choice = st.selectbox( "Choose a model for prediction", ("TCIP-Net ConvGRU", "TCIP-Net ConvLSTM", "TCIP-Net Traj-GRU", "TCIP-Net 3DCNN", "TCIP-Net Spatio-temporal LSTM", "TCUWSP-Net (Proposed Model)"), index=0 ) # Center-aligned, more attractive submit button st.markdown("
", unsafe_allow_html=True) col_btn1, col_btn2 = st.columns(2) with col_btn1: submit_button = st.button("Predict Intensity", use_container_width=True) with col_btn2: all_models_button = st.button("Predict Intensity for All Models", use_container_width=True) # ------------------ Process Single Model Button ------------------ if submit_button: if len(ir_images) == 8 and len(pmw_images) == 8: # st.success("Starting preprocessing...") if model_choice == "TCUWSP-Net (Proposed Model)": from unetlstm import predict_unetlstm model_predict_fn = predict_unetlstm elif model_choice == "TCIP-Net ConvGRU": from gru_model import predict model_predict_fn = predict elif model_choice == "TCIP-Net ConvLSTM": from convlstm import predict_lstm model_predict_fn = predict_lstm elif model_choice == "TCIP-Net 3DCNN": from cnn3d import predict_3dcnn model_predict_fn = predict_3dcnn elif model_choice == "TCIP-Net Traj-GRU": from trjgru import predict_trajgru model_predict_fn = predict_trajgru elif model_choice == "TCIP-Net Spatio-temporal LSTM": from spaio_temp import predict_stlstm model_predict_fn = predict_stlstm ir_arrays = [] pmw_arrays = [] train_vmax_2d = reshape_vmax(np.array(vmax_values)) train_vmax_3d= create_3d_vmax(train_vmax_2d) lat_processed = process_lat_values(lat_values) lon_processed = process_lon_values(lon_values) v_max_diff = calculate_intensity_difference(train_vmax_2d) for ir in ir_images: img = Image.open(ir).convert("L") arr = np.array(img).astype(np.float32) bt_arr = (arr / 255.0) * (310 - 190) + 190 resized = cv2.resize(bt_arr, (95, 95), interpolation=cv2.INTER_CUBIC) ir_arrays.append(resized) for pmw in pmw_images: img = Image.open(pmw).convert("L") arr = np.array(img).astype(np.float32) / 255.0 resized = cv2.resize(arr, (95, 95), interpolation=cv2.INTER_CUBIC) pmw_arrays.append(resized) ir=np.array(ir_arrays) pmw=np.array(pmw_arrays) # Stack into (8, 95, 95) ir_seq = process_images(ir) pmw_seq = process_images(pmw) # For demonstration: create batches X_train_new = ir_seq.reshape((1, 8, 95, 95)) # Shape: (1, 8, 95, 95) cc_mask= compute_convective_core_masks(X_train_new) hov_m_train = generate_hovmoller(X_train_new) hov_m_train[np.isnan(hov_m_train)] = 0 hov_m_train = hov_m_train.transpose(0, 2, 3, 1) cc_mask[np.isnan(cc_mask)] = 0 cc_mask=cc_mask.reshape(1, 8, 95, 95, 1) i_images=cc_mask+ir_seq reduced_images = np.concatenate([i_images,pmw_seq ], axis=-1) reduced_images[np.isnan(reduced_images)] = 0 if model_choice == "Unet_LSTM": import tensorflow as tf def tf_gradient_magnitude(images): # Sobel kernels sobel_x = tf.constant([[1, 0, -1], [2, 0, -2], [1, 0, -1]], dtype=tf.float32) sobel_y = tf.constant([[1, 2, 1], [0, 0, 0], [-1, -2, -1]], dtype=tf.float32) sobel_x = tf.reshape(sobel_x, [3, 3, 1, 1]) sobel_y = tf.reshape(sobel_y, [3, 3, 1, 1]) images = tf.convert_to_tensor(images, dtype=tf.float32) images = tf.expand_dims(images, -1) gx = tf.nn.conv2d(images, sobel_x, strides=1, padding='SAME') gy = tf.nn.conv2d(images, sobel_y, strides=1, padding='SAME') grad_mag = tf.sqrt(tf.square(gx) + tf.square(gy) + 1e-6) return tf.squeeze(grad_mag, -1).numpy() def GM_maps_prep(ir): GM_maps=[] for i in ir: GM_map = tf_gradient_magnitude(i) GM_maps.append(GM_map) GM_maps=np.array(GM_maps) return GM_maps ir_seq=ir_seq.reshape(8, 95, 95, 1) GM_maps = GM_maps_prep(ir_seq) print(GM_maps.shape) GM_maps=GM_maps.reshape(1, 8, 95, 95, 1) i_images=cc_mask+ir_seq+GM_maps reduced_images = np.concatenate([i_images,pmw_seq ], axis=-1) reduced_images[np.isnan(reduced_images)] = 0 print(reduced_images.shape) y = model_predict_fn(reduced_images, hov_m_train, train_vmax_3d, lat_processed, lon_processed, v_max_diff) else: y = model_predict_fn(reduced_images, hov_m_train, train_vmax_3d, lat_processed, lon_processed, v_max_diff) st.write("Predicted Maximum Sustained Wind Speed [Vmax] (in knots):", y) else: st.error("Make sure you uploaded exactly 8 IR and 8 PMW images.") # ------------------ Process All Models Button ------------------ if all_models_button: if len(ir_images) == 8 and len(pmw_images) == 8: st.info("Running predictions for all models... This may take a moment.") # Store all model names and prediction functions all_model_names = [ "TCIP-Net (3DCNN)", "TCIP-Net (ST-LSTM)", "TCIP-Net (ConvLSTM)", "TCIP-Net (TrajGRU)", "TCIP-Net (ConvGRU)", "TCUWSP-Net (Proposed)" ] # Process input data once for all models ir_arrays = [] pmw_arrays = [] train_vmax_2d = reshape_vmax(np.array(vmax_values)) train_vmax_3d = create_3d_vmax(train_vmax_2d) lat_processed = process_lat_values(lat_values) lon_processed = process_lon_values(lon_values) v_max_diff = calculate_intensity_difference(train_vmax_2d) for ir in ir_images: img = Image.open(ir).convert("L") arr = np.array(img).astype(np.float32) bt_arr = (arr / 255.0) * (310 - 190) + 190 resized = cv2.resize(bt_arr, (95, 95), interpolation=cv2.INTER_CUBIC) ir_arrays.append(resized) for pmw in pmw_images: img = Image.open(pmw).convert("L") arr = np.array(img).astype(np.float32) / 255.0 resized = cv2.resize(arr, (95, 95), interpolation=cv2.INTER_CUBIC) pmw_arrays.append(resized) ir = np.array(ir_arrays) pmw = np.array(pmw_arrays) ir_seq = process_images(ir) pmw_seq = process_images(pmw) X_train_new = ir_seq.reshape((1, 8, 95, 95)) cc_mask = compute_convective_core_masks(X_train_new) hov_m_train = generate_hovmoller(X_train_new) hov_m_train[np.isnan(hov_m_train)] = 0 hov_m_train = hov_m_train.transpose(0, 2, 3, 1) cc_mask[np.isnan(cc_mask)] = 0 cc_mask = cc_mask.reshape(1, 8, 95, 95, 1) i_images = cc_mask + ir_seq reduced_images = np.concatenate([i_images, pmw_seq], axis=-1) reduced_images[np.isnan(reduced_images)] = 0 # Special processing for Unet_LSTM model if needed import tensorflow as tf def tf_gradient_magnitude(images): # Sobel kernels sobel_x = tf.constant([[1, 0, -1], [2, 0, -2], [1, 0, -1]], dtype=tf.float32) sobel_y = tf.constant([[1, 2, 1], [0, 0, 0], [-1, -2, -1]], dtype=tf.float32) sobel_x = tf.reshape(sobel_x, [3, 3, 1, 1]) sobel_y = tf.reshape(sobel_y, [3, 3, 1, 1]) images = tf.convert_to_tensor(images, dtype=tf.float32) images = tf.expand_dims(images, -1) gx = tf.nn.conv2d(images, sobel_x, strides=1, padding='SAME') gy = tf.nn.conv2d(images, sobel_y, strides=1, padding='SAME') grad_mag = tf.sqrt(tf.square(gx) + tf.square(gy) + 1e-6) return tf.squeeze(grad_mag, -1).numpy() def GM_maps_prep(ir): GM_maps=[] for i in ir: GM_map = tf_gradient_magnitude(i) GM_maps.append(GM_map) GM_maps=np.array(GM_maps) return GM_maps # For Unet_LSTM model ir_seq_reshaped = ir_seq.reshape(8, 95, 95, 1) GM_maps = GM_maps_prep(ir_seq_reshaped) GM_maps = GM_maps.reshape(1, 8, 95, 95, 1) i_images_unet = cc_mask + ir_seq_reshaped + GM_maps reduced_images_unet = np.concatenate([i_images_unet, pmw_seq], axis=-1) reduced_images_unet[np.isnan(reduced_images_unet)] = 0 # Run predictions for all models predictions = [] progress_bar = st.progress(0) # Import all prediction functions from cnn3d import predict_3dcnn from spaio_temp import predict_stlstm from convlstm import predict_lstm from trjgru import predict_trajgru from gru_model import predict from unetlstm import predict_unetlstm prediction_functions = [ predict_3dcnn, # 3DCNN predict_stlstm, # ST-LSTM predict_lstm, # ConvLSTM predict_trajgru, # TrajGRU predict, # ConvGRU predict_unetlstm # TCUWSP-Net ] # Run predictions for i, predict_fn in enumerate(prediction_functions): progress_bar.progress((i) / len(prediction_functions)) # Special case for TCUWSP-Net (Proposed Model) if i == 5: # TCUWSP-Net index y = predict_fn(reduced_images_unet, hov_m_train, train_vmax_3d, lat_processed, lon_processed, v_max_diff) else: y = predict_fn(reduced_images, hov_m_train, train_vmax_3d, lat_processed, lon_processed, v_max_diff) predictions.append(float(y)) progress_bar.progress(1.0) # Create results DataFrame results_data = { "Model": all_model_names, "RMSE": ablation_data["RMSE"], "MAE": ablation_data["MAE"], "Predicted Vmax (kt)": predictions } results_df = pd.DataFrame(results_data) # Show DataFrame st.subheader("Prediction Results from All Models") st.dataframe(results_df, use_container_width=True) # Generate and display comparison chart st.subheader("Visual Comparison of Models") # Prepare data for visualization plot_model_names = [name.replace(" ", "\n") for name in all_model_names] mae_values = results_df["MAE"].tolist() rmse_values = results_df["RMSE"].tolist() predicted_values = results_df["Predicted Vmax (kt)"].tolist() # Generate figure fig = generate_comparison_chart(plot_model_names, mae_values, rmse_values, predicted_values) # Display figure st.pyplot(fig) # Add some interpretation st.subheader("Interpretation") st.write(""" - **RMSE and MAE**: Lower values indicate better model performance. - **Percentage Improvements**: Show reduction in error compared to the baseline TCIP-Net (3DCNN) model. - **Predicted Vmax**: The current intensity prediction for the tropical cyclone based on the provided imagery and historical data. """) # Highlight best model best_model_idx = rmse_values.index(min(rmse_values)) best_model = all_model_names[best_model_idx] best_prediction = predicted_values[best_model_idx] st.success(f"🌟 Best performing model: **{best_model}** with RMSE: **{min(rmse_values):.2f} kt** and predicted intensity: **{best_prediction:.2f} kt**") else: st.error("Make sure you uploaded exactly 8 IR and 8 PMW images.")