Update app.py

app.py

@@ -6,14 +6,19 @@ import numpy as np
 import torchvision.transforms as transforms
 import torch
 from utils import *
-#from utils import flip_odd_lines, modulo, center_modulo, unmodulo, hard_thresholding, stripe_estimation, recons
 from utils import modulo
 import cv2
-
 import matplotlib.pyplot as plt
 
-def process_image(image, model_id, image_size, conf_threshold, correction, sat_factor, kernel_size, DO, t, vertical):
+# Definir el tamaño de la imagen constante
+IMAGE_SIZE = 640  # Asignar el valor constante para image_size
 
+def process_image(image, model_id, sat_factor, DO, t, vertical):
+    # Definir valores fijos dentro de la función
+    conf_threshold = 0.85
+    correction = 1.0
+    kernel_size = 7  # Definir kernel_size como constante
+    
     original_image = np.array(image)
     original_image = original_image - original_image.min()
     original_image = original_image / original_image.max()
@@ -24,39 +29,35 @@ def process_image(image, model_id, image_size, conf_threshold, correction, sat_f
     scaling = 1.0
     original_image = cv2.resize(original_image, (0, 0), fx=scaling, fy=scaling)
 
-
     blurred_image = apply_blur(original_image / 255.0, kernel_size)
     clipped_image = clip_image(blurred_image, correction, sat_factor) 
 
-    img_tensor = torch.tensor(blurred_image, dtype=torch.float32 ).permute(2, 0, 1).unsqueeze(0)
-    img_tensor = modulo( img_tensor * sat_factor, L=1.0)
+    img_tensor = torch.tensor(blurred_image, dtype=torch.float32).permute(2, 0, 1).unsqueeze(0)
+    img_tensor = modulo(img_tensor * sat_factor, L=1.0)
 
     wrapped_image = img_tensor.squeeze(0).permute(1, 2, 0).numpy()
     wrapped_image = (wrapped_image*255).astype(np.uint8)
 
-    original_annotated, original_detections = yolov10_inference(original_image, model_id, image_size, conf_threshold)
-    clipped_annotated, clipped_detections = yolov10_inference((clipped_image*255.0).astype(np.uint8), "yolov10n", image_size, conf_threshold)
-    wrapped_annotated, wrapped_detections = yolov10_inference(wrapped_image, model_id, image_size, conf_threshold)
+    original_annotated, original_detections = yolov10_inference(original_image, model_id, IMAGE_SIZE, conf_threshold)
+    clipped_annotated, clipped_detections = yolov10_inference((clipped_image*255.0).astype(np.uint8), "yolov10n", IMAGE_SIZE, conf_threshold)
+    wrapped_annotated, wrapped_detections = yolov10_inference(wrapped_image, model_id, IMAGE_SIZE, conf_threshold)
 
-    # Assuming `recons` is a function in `utils.py`
     recon_image = recons(img_tensor, DO=1, L=1.0, vertical=(vertical == "True"), t=t)
     recon_image_pil = transforms.ToPILImage()(recon_image.squeeze(0))
     recon_image_np = np.array(recon_image_pil).astype(np.uint8)
 
+    recon_annotated, recon_detections = yolov10_inference(recon_image_np, model_id, IMAGE_SIZE, conf_threshold)
 
-    recon_annotated, recon_detections = yolov10_inference(recon_image_np, model_id, image_size, conf_threshold)
-
-
-
-    return original_annotated, clipped_annotated, wrapped_annotated, recon_annotated, 
-
+    metrics_clip = calculate_detection_metrics(original_detections, clipped_detections)
+    metrics_wrap = calculate_detection_metrics(original_detections, wrapped_detections)
+    metrics_recons = calculate_detection_metrics(original_detections, recon_detections)
 
+    return original_annotated, clipped_annotated, wrapped_annotated, recon_annotated, metrics_clip, metrics_wrap, metrics_recons
 
 def app():
-    # Variables para ajustar el tamaño de las imágenes
-    image_scaler = 0.55  # Cambia este valor para ajustar el tamaño de las imágenes
-    image_width = int(600 * image_scaler)  # Ancho de la imagen
-    image_height = int(200 * image_scaler)  # Alto de la imagen
+    image_scaler = 0.55  
+    image_width = int(600 * image_scaler)  
+    image_height = int(200 * image_scaler)
 
     with gr.Blocks(css=f"""
     .fixed-size-image img {{
@@ -72,7 +73,7 @@ def app():
     .gr-column {{
         display: flex;
         flex-direction: column;
-        align-items: center;  /* Centramos todo en la columna */
+        align-items: center;
         padding: 0 !important;
         margin: 0 !important;
     }}
@@ -101,40 +102,31 @@ def app():
         background-color: #e94e42;
     }}
     """) as demo:
-        # Título completamente centrado
         gr.Markdown("## Modulo Imaging for Computer Vision", elem_id="centered-title")
 
-        # Organizar los artículos en dos columnas con botones centrados bajo el título
         with gr.Row():
-            # Primer artículo con su botón centrado en su columna
             with gr.Column():
                 gr.Markdown("### High Dynamic Range Modulo Imaging for Robust Object Detection in Autonomous Driving", elem_id="centered-text")
                 gr.HTML('<a href="https://openreview.net/pdf?id=2GqZFx2I7s" target="_blank" class="custom-button btn-grey">Article</a>')
             
-            # Segundo artículo con su botón centrado en su columna
             with gr.Column():
                 gr.Markdown("### Autoregressive High-Order Finite Difference Modulo Imaging: High-Dynamic Range for Computer Vision Applications", elem_id="centered-text")
                 gr.HTML('<a href="https://cvlai.net/aim/2024/" target="_blank" class="custom-button btn-red">Article</a>')
 
-        # Ajustes de sliders y controles eliminados, valores fijos
         image = gr.Image(type="pil", label="Upload Image")
         model_id = gr.Dropdown(label="Model", choices=["yolov10n", "yolov10s", "yolov10m", "yolov10b", "yolov10l", "yolov10x"], value="yolov10x")
         
-        # Valores fijos
-        conf_threshold = 0.85
-        correction = 1.0
-        kernel_size = 7
-        DO = "1"
-        image_size = 640
-        t = 0.7
-        vertical = "True"
-
-        # Mantener el control del nivel de saturación
         sat_factor = gr.Slider(label="Saturation Factor", minimum=1.0, maximum=5.0, step=0.1, value=2.0)
+        DO = gr.Radio(label="DO", choices=["1", "2"], value="1")
+        t = gr.Slider(label="t", minimum=0.0, maximum=1.0, step=0.1, value=0.7)
+        vertical = gr.Radio(label="Vertical", choices=["True", "False"], value="True")
 
+        with gr.Row():
+            spud_button = gr.Button("SPUD", elem_id="spud-button")
+            ahfd_button = gr.Button("AHFD", elem_id="ahfd-button")
+        
         process_button = gr.Button("Process Image")
 
-        # Mostrar las imágenes en dos columnas, dos imágenes por columna
         with gr.Row():
             with gr.Column():
                 output_original = gr.Image(label="Original + blur")
@@ -143,56 +135,14 @@ def app():
                 output_wrap = gr.Image(label="Wrapped")
                 output_recons = gr.Image(label="Reconstructed")
 
-
-
-
-
         process_button.click(
             fn=process_image,
-            inputs=[image, model_id, image_size, conf_threshold, correction, sat_factor, kernel_size, DO, t, vertical],
+            inputs=[image, model_id, sat_factor, DO, t, vertical],
             outputs=[output_original, output_clip, output_wrap, output_recons]
         )
 
-
-
-
-
-
-
-
-
-
-
-        
-
-        # Añadir botón que despliega cómo citar el artículo en formato BibTeX
-        def show_bibtex_citation():
-            bibtex_citation = '''@inproceedings{contreras2024modulo,
-  title={High Dynamic Range Modulo Imaging for Robust Object Detection in Autonomous Driving},
-  author={Contreras, Kebin and Monroy, Brayan and Bacca, Jorge},
-  booktitle={Autonomous Driving Conference},
-  year={2024},
-  address={Bucaramanga, Colombia},
-  publisher={Universidad Industrial de Santander}
-}'''
-            return bibtex_citation
-
-        # Añadir botón para copiar el texto
-        citation_button = gr.Button("How to Cite this Article (BibTeX)")
-        citation_text = gr.Textbox(visible=False, label="BibTeX Citation", interactive=False)
-        copy_button = gr.Button("Copy Citation", visible=False)
-
-        # Mostrar la cita en formato BibTeX y habilitar botón de copiar
-        def copy_citation(citation):
-            return gr.update(visible=True), citation
-
-        citation_button.click(show_bibtex_citation, outputs=[citation_text])
-        copy_button.click(lambda x: x, inputs=citation_text, outputs=citation_text)
-
-    
-
     return demo
 
-
 if __name__ == "__main__":
     app().launch()
+