Upload functional_wlan_design_gradio.py

functional_wlan_design_gradio.py

@@ -0,0 +1,372 @@
+# -*- coding: utf-8 -*-
+"""Funcional WLAN_design_gradio.ipynb
+
+Automatically generated by Colab.
+
+Original file is located at
+    https://colab.research.google.com/drive/1MIfY3UkK4eSXOiPx3gtMSPoZMtnyJxux
+"""
+
+from google.colab import drive
+drive.mount('/content/drive')
+
+# Commented out IPython magic to ensure Python compatibility.
+# %%capture
+# !pip install gradio
+
+import gradio as gr
+from PIL import Image
+import os
+from tensorflow.keras.models import load_model
+import numpy as np
+import matplotlib.pyplot as plt
+from matplotlib.colors import Normalize
+from io import BytesIO
+
+# Images path
+path_main = '/content/drive/Othercomputers/False-2-Tesis-Maestria /Phase-3-thesis/'
+images_file = path_main + 'Scennarios init/Scennarios W'
+
+# Load DL models
+modelo_1ap = load_model(path_main + 'Models/SINR-2APs/modelo_1ap_app.keras')
+modelo_2ap = load_model(path_main + 'Models/SINR-2APs/modelo_2ap_app.keras')
+
+plt.rc('font', family='Times New Roman')
+fontsize_t = 15
+
+def coordinates_process(texto):
+    coordinates = texto.split("), ")
+
+    resultado = []
+    for coord in coordinates:
+        try:
+            coord = coord.replace("(", "").replace(")", "")
+            x, y = map(int, coord.split(","))
+            # Validate range
+            if 0 <= x <= 255 and 0 <= y <= 255:
+                resultado.append((x, y))
+            else:
+                return False
+        except ValueError:
+            return False
+
+    while len(resultado) < 3:
+        resultado.append((0, 0))
+
+    return resultado
+
+# plan images path
+def plan_images_list():
+    return [file_ for file_ in os.listdir(images_file) if file_.endswith((".JPG", ".jpg", ".jpeg", ".png"))]
+
+# Valdate inputs
+def validate_input(value):
+    if value == "" or value is None:
+        return 0
+    elif value >= 0 or value <= 2:
+        return value
+
+# MAIN FUNCTION ****************************************************************
+def main_function(plan_name, apch1, apch6, apch11, coord1, coord6, coord11):
+    image_plan_path = os.path.join(images_file, plan_name)
+    imagen1 = Image.open(image_plan_path)
+
+    # No negative number as input
+    if not (0 <= apch1 <= 2):
+        return False
+    if not (0 <= apch6 <= 2):
+        return False
+    if not (0 <= apch11 <= 2):
+        return False
+
+    # Some variables init
+    deep_count = 0
+    deep_coverage = []
+    channels = [1, 6, 11]
+    num_APs = np.zeros(len(channels), dtype=int)
+    num_APs[0] = apch1
+    num_APs[1] = apch6
+    num_APs[2] = apch11
+    dimension = 256
+    aps_chs = np.zeros((dimension, dimension, len(channels)))
+
+    # Load plan
+    numero = plan_name[:1]
+    plan_in = np.array(Image.open(f"{path_main}Scennarios init/Scennarios B/{numero}.png")) / 255
+
+    coords = [coord1, coord6, coord11]
+    for att, channel in enumerate(channels):
+      if num_APs[att] > 0:
+          coordinates = coordinates_process(coords[att])
+          for x, y in coordinates:
+            if x != 0 and y != 0:
+              aps_chs[int(y), int(x), att] = 1
+
+    # Coverage process
+    deep_coverage = []
+    ap_images = []
+    layer_indices = []
+    imagencober = {}
+    for k in range(len(channels)):
+        capa = aps_chs[:, :, k]
+        filas, columnas = np.where(capa == 1)
+
+        if len(filas) == 2:
+            # For 2 AP
+            deep_count += 1
+            layer_1 = np.zeros_like(capa)
+            layer_2 = np.zeros_like(capa)
+            layer_1[filas[0], columnas[0]] = 1
+            layer_2[filas[1], columnas[1]] = 1
+
+            datos_entrada = np.stack([plan_in, layer_1, layer_2], axis=-1)
+            prediction = modelo_2ap.predict(datos_entrada[np.newaxis, ...])[0]
+
+        elif len(filas) == 1:
+            # For 1 AP
+            deep_count += 1
+            layer_1 = np.zeros_like(capa)
+            layer_1[filas[0], columnas[0]] = 1
+
+            datos_entrada = np.stack([plan_in, layer_1], axis=-1)
+            prediction = modelo_1ap.predict(datos_entrada[np.newaxis, ...])[0]
+
+        else:
+            # Whitout AP
+            prediction = np.zeros((dimension,dimension,1))
+
+        # print(prediction.shape)
+        deep_coverage.append(prediction)
+        prediction_rgb = np.squeeze((Normalize()(prediction)))
+        ap_images.append(prediction_rgb)  # Guardar la imagen de cobertura del AP
+
+        if np.all(prediction == 0):
+            plt.imshow(prediction_rgb)
+            plt.title('No coverage', fontsize=fontsize_t + 2, family='Times New Roman')
+            plt.axis("off")
+        else:
+            plt.imshow(prediction_rgb, cmap='jet')
+            cbar = plt.colorbar(ticks=np.linspace(0, 1, num=6),)
+            cbar.set_label('SINR [dB]', fontsize=fontsize_t, fontname='Times New Roman')
+            cbar.set_ticklabels(['-3.01', '20.29', '43.60', '66.90', '90.20', '113.51'])
+            cbar.ax.tick_params(labelsize=fontsize_t, labelfontfamily = 'Times New Roman')
+            plt.axis("off")
+
+        # Save the plot to a buffer
+        buf = BytesIO()
+        plt.savefig(buf, format='png')
+        buf.seek(0)
+        plt.close()
+
+        # Convert buffer to an image
+        imagencober[k] = Image.open(buf)
+
+        # Cell map estimation
+        layer_indices.append(np.argmax(prediction, axis=0))
+
+    # Final coverage
+    if deep_coverage:
+        deep_coverage = np.array(deep_coverage)
+        nor_matrix = np.max(deep_coverage, axis=0)
+        celdas = np.argmax(deep_coverage, axis=0)
+
+        resultado_rgb = np.squeeze((Normalize()(nor_matrix)))
+
+        plt.imshow(resultado_rgb, cmap='jet')
+        cbar = plt.colorbar(ticks=np.linspace(0, 1, num=6))
+        cbar.set_label('SINR [dB]', fontsize=fontsize_t, fontname='Times New Roman')
+        cbar.set_ticklabels(['-3.01', '20.29', '43.60', '66.90', '90.20', '113.51'])
+        cbar.ax.tick_params(labelsize=fontsize_t, labelfontfamily = 'Times New Roman')
+        plt.axis("off")
+
+        # Save the plot to a buffer
+        buf = BytesIO()
+        plt.savefig(buf, format='png')
+        buf.seek(0)
+        plt.close()
+
+        # Convert buffer to an image
+        imagen3 = Image.open(buf)
+
+        if num_APs[0] > 0 and num_APs[1] > 0 and num_APs[2] > 0:
+            cmap = plt.cm.colors.ListedColormap(['blue', 'red', 'green'])
+            plt.imshow(celdas, cmap=cmap)
+            cbar = plt.colorbar()
+            cbar.set_ticks([0, 1, 2])
+            cbar.set_ticklabels(['1', '6', '11'])
+            cbar.set_label('Cell ID', fontsize=fontsize_t, fontname='Times New Roman')
+            cbar.ax.tick_params(labelsize=fontsize_t, labelfontfamily = 'Times New Roman')
+            plt.axis("off")
+
+            # Save the plot to a buffer
+            buf = BytesIO()
+            plt.savefig(buf, format='png')
+            buf.seek(0)
+            plt.close()
+
+            # Convert buffer to an image
+            imagen4 = Image.open(buf)
+
+        elif num_APs[0] > 0 and num_APs[1] > 0:
+            cmap = plt.cm.colors.ListedColormap(['blue', 'red'])
+            plt.imshow(celdas, cmap=cmap)
+            cbar = plt.colorbar()
+            cbar.set_ticks([0, 1])
+            cbar.set_ticklabels(['1', '6'])
+            cbar.set_label('Cell ID', fontsize=fontsize_t, fontname='Times New Roman')
+            cbar.ax.tick_params(labelsize=fontsize_t, labelfontfamily = 'Times New Roman')
+            plt.axis("off")
+
+            # Save the plot to a buffer
+            buf = BytesIO()
+            plt.savefig(buf, format='png')
+            buf.seek(0)
+            plt.close()
+
+            # Convert buffer to an image
+            imagen4 = Image.open(buf)
+
+        elif num_APs[0] > 0 and num_APs[2] > 0:
+            cmap = plt.cm.colors.ListedColormap(['blue', 'red'])
+            plt.imshow(celdas, cmap=cmap)
+            cbar = plt.colorbar()
+            cbar.set_ticks([0, 1])
+            cbar.set_ticklabels(['1', '11'])
+            cbar.set_label('Cell ID', fontsize=fontsize_t, fontname='Times New Roman')
+            cbar.ax.tick_params(labelsize=fontsize_t, labelfontfamily = 'Times New Roman')
+            plt.axis("off")
+
+            # Save the plot to a buffer
+            buf = BytesIO()
+            plt.savefig(buf, format='png')
+            buf.seek(0)
+            plt.close()
+
+            # Convert buffer to an image
+            imagen4 = Image.open(buf)
+
+        elif num_APs[1] > 0 and num_APs[2] > 0:
+            cmap = plt.cm.colors.ListedColormap(['blue', 'red'])
+            plt.imshow(celdas, cmap=cmap)
+            cbar = plt.colorbar()
+            cbar.set_ticks([0, 1])
+            cbar.set_ticklabels(['6', '11'])
+            cbar.set_label('Cell ID', fontsize=fontsize_t, fontname='Times New Roman')
+            cbar.ax.tick_params(labelsize=fontsize_t, labelfontfamily = 'Times New Roman')
+            plt.axis("off")
+
+            # Save the plot to a buffer
+            buf = BytesIO()
+            plt.savefig(buf, format='png')
+            buf.seek(0)
+            plt.close()
+
+            # Convert buffer to an image
+            imagen4 = Image.open(buf)
+
+        else:
+            cmap = plt.cm.colors.ListedColormap(['blue'])
+            plt.imshow(celdas, cmap=cmap)
+            cbar = plt.colorbar()
+            cbar.set_ticks([0])
+            cbar.set_ticklabels(['1'])
+            cbar.set_label('Cell ID', fontsize=fontsize_t, fontname='Times New Roman')
+            cbar.ax.tick_params(labelsize=fontsize_t, labelfontfamily = 'Times New Roman')
+            plt.axis("off")
+
+            # Save the plot to a buffer
+            buf = BytesIO()
+            plt.savefig(buf, format='png')
+            buf.seek(0)
+            plt.close()
+
+            # Convert buffer to an image
+            imagen4 = Image.open(buf)
+
+    return [imagencober[0], imagencober[1], imagencober[2], imagen3, imagen4]
+
+# plan visualization
+def load_plan_vi(mapa_seleccionado):
+
+    image_plan_path1 = os.path.join(images_file, mapa_seleccionado)
+    plan_image = Image.open(image_plan_path1)
+
+    plan_n = np.array(plan_image.convert('RGB'))
+    plt.figure(figsize=(3, 3))
+    plt.imshow(plan_n)
+    plt.xticks(np.arange(0, 256, 50))
+    plt.yticks(np.arange(0, 256, 50))
+
+    # Save the plot to a buffer
+    buf = BytesIO()
+    plt.savefig(buf, format='png')
+    buf.seek(0)
+    plt.close()
+
+    # Convert buffer to an image
+    plan_im = Image.open(buf)
+
+    return plan_im
+
+with gr.Blocks() as demo:
+
+    gr.Markdown("""
+    ## Fast Radio Propagation Prediction in WLANs Using Deep Learning
+    This app use deep learning models to radio map estimation (RME). RME entails estimating the received RF power based on spatial information maps.
+
+    Instructions for use:
+
+    - A predefined list of indoor floor plans is available for use.
+    - A maximum of (255,255) pixels is allowed for image size.
+    - Negative numbers are not allowed.
+    - The established format for the coordinates of each access point (AP) must be maintained.
+    - A maximum of 2 APs are allowed per channel.
+    """)
+
+    with gr.Row():
+        # Input left panel
+        with gr.Column(scale=1):  # Scale to resize
+            map_dropdown = gr.Dropdown(choices=plan_images_list(), label="Select indoor plan")
+            ch1_input = gr.Number(label="APs CH 1")
+            ch6_input = gr.Number(label="APs CH 6")
+            ch11_input = gr.Number(label="APs CH 11")
+            coords_ch1_input = gr.Textbox(label="Coordinate CH 1", placeholder="Format: (x1, y1), (x2, y2)")
+            coords_ch6_input = gr.Textbox(label="Coordinate CH 6", placeholder="Format: (x1, y1), (x2, y2)")
+            coords_ch11_input = gr.Textbox(label="Coordinate CH 11", placeholder="Format: (x1, y1), (x2, y2)")
+            button1 = gr.Button("Load plan")
+            button2 = gr.Button("Predict coverage")
+
+        # Rigth panel
+        a_images = 320 # Size putput images
+
+        with gr.Column(scale=3):
+            first_image_output = gr.Image(label="plan image", height=a_images, width=a_images)
+            # with gr.Row():
+            with gr.Row():
+                image_ch1 = gr.Image(label="CH 1 coverage", height=a_images, width=a_images)
+                image_ch6 = gr.Image(label="CH 6 coverage", height=a_images, width=a_images)
+                image_ch11 = gr.Image(label="CH 11 coverage", height=a_images, width=a_images)
+
+            with gr.Row():
+                image_cover_final = gr.Image(label="Final coverage", height=a_images, width=a_images)
+                image_cells = gr.Image(label="Cells coverage", height=a_images, width=a_images)
+
+
+    # Buttons
+    button1.click(load_plan_vi, inputs=[map_dropdown], outputs=first_image_output)
+    button2.click(
+    lambda map_dropdown, ch1, ch6, ch11, coords1, coords6, coords11: main_function(
+        map_dropdown,
+        validate_input(ch1),
+        validate_input(ch6),
+        validate_input(ch11),
+        coords1,
+        coords6,
+        coords11,
+    ),
+    inputs=[map_dropdown, ch1_input, ch6_input, ch11_input, coords_ch1_input, coords_ch6_input, coords_ch11_input],
+    outputs=[image_ch1, image_ch6, image_ch11, image_cover_final, image_cells]
+)
+
+
+demo.launch()