wi-lab commited on
Commit
c82ad5f
·
verified ·
1 Parent(s): 5f28bed

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +18 -29
app.py CHANGED
@@ -15,7 +15,7 @@ from sklearn.metrics import f1_score
15
  import seaborn as sns
16
 
17
  #################### BEAM PREDICTION #########################}
18
- def beam_prediction_task(data_percentage, task_complexity):
19
  # Folder naming convention based on input_type, data_percentage, and task_complexity
20
  raw_folder = f"images/raw_{data_percentage/100:.1f}_{task_complexity}"
21
  embeddings_folder = f"images/embedding_{data_percentage/100:.1f}_{task_complexity}"
@@ -24,7 +24,7 @@ def beam_prediction_task(data_percentage, task_complexity):
24
  raw_cm = compute_average_confusion_matrix(raw_folder)
25
  if raw_cm is not None:
26
  raw_cm_path = os.path.join(raw_folder, "confusion_matrix_raw.png")
27
- plot_confusion_matrix_beamPred(raw_cm, classes=np.arange(raw_cm.shape[0]), title=f"Raw Confusion Matrix\n({data_percentage}% data, {task_complexity} beams)", save_path=raw_cm_path)
28
  raw_img = Image.open(raw_cm_path)
29
  else:
30
  raw_img = None
@@ -33,15 +33,13 @@ def beam_prediction_task(data_percentage, task_complexity):
33
  embeddings_cm = compute_average_confusion_matrix(embeddings_folder)
34
  if embeddings_cm is not None:
35
  embeddings_cm_path = os.path.join(embeddings_folder, "confusion_matrix_embeddings.png")
36
- plot_confusion_matrix_beamPred(embeddings_cm, classes=np.arange(embeddings_cm.shape[0]), title=f"Embeddings Confusion Matrix\n({data_percentage}% data, {task_complexity} beams)", save_path=embeddings_cm_path)
37
  embeddings_img = Image.open(embeddings_cm_path)
38
  else:
39
  embeddings_img = None
40
 
41
  return raw_img, embeddings_img
42
 
43
- from sklearn.metrics import f1_score
44
-
45
  # Function to compute the F1-score based on the confusion matrix
46
  def compute_f1_score(cm):
47
  # Compute precision and recall
@@ -61,30 +59,12 @@ def compute_f1_score(cm):
61
  f1 = np.nan_to_num(f1) # Replace NaN with 0
62
  return np.mean(f1) # Return the mean F1-score across all classes
63
 
64
- import matplotlib.pyplot as plt
65
- import seaborn as sns
66
- import numpy as np
67
- from PIL import Image
68
-
69
- def plot_confusion_matrix_beamPred(cm, classes, title, save_path, dark_mode=None):
70
- """
71
- Plot confusion matrix and adjust colors based on light/dark mode settings.
72
- :param cm: Confusion matrix data.
73
- :param classes: List of class labels.
74
- :param title: Plot title.
75
- :param save_path: Path to save the plot.
76
- :param dark_mode: Boolean to toggle between light and dark modes. If None, use the current theme.
77
- """
78
-
79
- # If dark_mode is None, try detecting it from rcParams (matplotlib theme)
80
- if dark_mode is None:
81
- dark_mode = plt.rcParams['axes.facecolor'] == '#333333' # Check if dark background is set
82
-
83
  # Compute the average F1-score
84
  avg_f1 = compute_f1_score(cm)
85
 
86
- # Choose the color scheme based on the mode
87
- if dark_mode:
88
  plt.style.use('dark_background') # Use dark mode styling
89
  text_color = 'white'
90
  cmap = 'cividis' # Dark-mode-friendly colormap
@@ -95,7 +75,7 @@ def plot_confusion_matrix_beamPred(cm, classes, title, save_path, dark_mode=None
95
 
96
  plt.figure(figsize=(10, 10))
97
 
98
- # Plot the confusion matrix with the selected colormap
99
  sns.heatmap(cm, cmap=cmap, cbar=True, linecolor='white', vmin=0, vmax=cm.max(), alpha=0.85)
100
 
101
  # Add F1-score to the title
@@ -579,12 +559,21 @@ with gr.Blocks(css="""
579
  task_complexity_dropdown = gr.Dropdown(label="Task Complexity (Number of Beams)", choices=[16, 32, 64, 128, 256], value=16)
580
 
581
  with gr.Row():
 
 
 
 
 
 
 
 
 
582
  raw_img_bp = gr.Image(label="Raw Channels", type="pil", width=300, height=500)
583
  embeddings_img_bp = gr.Image(label="Embeddings", type="pil", width=300, height=500)
584
 
585
  # Update the confusion matrices whenever sliders change
586
- data_percentage_slider.change(fn=beam_prediction_task, inputs=[data_percentage_slider, task_complexity_dropdown], outputs=[raw_img_bp, embeddings_img_bp])
587
- task_complexity_dropdown.change(fn=beam_prediction_task, inputs=[data_percentage_slider, task_complexity_dropdown], outputs=[raw_img_bp, embeddings_img_bp])
588
 
589
  # Add a conclusion section at the bottom
590
  gr.Markdown("""
 
15
  import seaborn as sns
16
 
17
  #################### BEAM PREDICTION #########################}
18
+ def beam_prediction_task(data_percentage, task_complexity, user_mode="light"):
19
  # Folder naming convention based on input_type, data_percentage, and task_complexity
20
  raw_folder = f"images/raw_{data_percentage/100:.1f}_{task_complexity}"
21
  embeddings_folder = f"images/embedding_{data_percentage/100:.1f}_{task_complexity}"
 
24
  raw_cm = compute_average_confusion_matrix(raw_folder)
25
  if raw_cm is not None:
26
  raw_cm_path = os.path.join(raw_folder, "confusion_matrix_raw.png")
27
+ plot_confusion_matrix_beamPred(raw_cm, classes=np.arange(raw_cm.shape[0]), title=f"Raw Confusion Matrix\n({data_percentage}% data, {task_complexity} beams)", save_path=raw_cm_path, user_mode=user_mode)
28
  raw_img = Image.open(raw_cm_path)
29
  else:
30
  raw_img = None
 
33
  embeddings_cm = compute_average_confusion_matrix(embeddings_folder)
34
  if embeddings_cm is not None:
35
  embeddings_cm_path = os.path.join(embeddings_folder, "confusion_matrix_embeddings.png")
36
+ plot_confusion_matrix_beamPred(embeddings_cm, classes=np.arange(embeddings_cm.shape[0]), title=f"Embeddings Confusion Matrix\n({data_percentage}% data, {task_complexity} beams)", save_path=embeddings_cm_path, user_mode=user_mode)
37
  embeddings_img = Image.open(embeddings_cm_path)
38
  else:
39
  embeddings_img = None
40
 
41
  return raw_img, embeddings_img
42
 
 
 
43
  # Function to compute the F1-score based on the confusion matrix
44
  def compute_f1_score(cm):
45
  # Compute precision and recall
 
59
  f1 = np.nan_to_num(f1) # Replace NaN with 0
60
  return np.mean(f1) # Return the mean F1-score across all classes
61
 
62
+ def plot_confusion_matrix_beamPred(cm, classes, title, save_path, user_mode="light"):
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
63
  # Compute the average F1-score
64
  avg_f1 = compute_f1_score(cm)
65
 
66
+ # Choose the color scheme based on the user's mode
67
+ if user_mode == 'dark':
68
  plt.style.use('dark_background') # Use dark mode styling
69
  text_color = 'white'
70
  cmap = 'cividis' # Dark-mode-friendly colormap
 
75
 
76
  plt.figure(figsize=(10, 10))
77
 
78
+ # Plot the confusion matrix with a colormap compatible for the mode
79
  sns.heatmap(cm, cmap=cmap, cbar=True, linecolor='white', vmin=0, vmax=cm.max(), alpha=0.85)
80
 
81
  # Add F1-score to the title
 
559
  task_complexity_dropdown = gr.Dropdown(label="Task Complexity (Number of Beams)", choices=[16, 32, 64, 128, 256], value=16)
560
 
561
  with gr.Row():
562
+
563
+ mode_input = gr.Textbox(visible=False) # Hidden input to capture user mode
564
+ gr.Markdown("""
565
+ <script>
566
+ const userPrefersDark = window.matchMedia && window.matchMedia('(prefers-color-scheme: dark)').matches;
567
+ document.querySelector('input[name="mode_input"]').value = userPrefersDark ? 'dark' : 'light';
568
+ </script>
569
+ """)
570
+
571
  raw_img_bp = gr.Image(label="Raw Channels", type="pil", width=300, height=500)
572
  embeddings_img_bp = gr.Image(label="Embeddings", type="pil", width=300, height=500)
573
 
574
  # Update the confusion matrices whenever sliders change
575
+ data_percentage_slider.change(fn=beam_prediction_task, inputs=[data_percentage_slider, task_complexity_dropdown, mode_input], outputs=[raw_img_bp, embeddings_img_bp])
576
+ task_complexity_dropdown.change(fn=beam_prediction_task, inputs=[data_percentage_slider, task_complexity_dropdown, mode_input], outputs=[raw_img_bp, embeddings_img_bp])
577
 
578
  # Add a conclusion section at the bottom
579
  gr.Markdown("""