valste commited on
Commit
bbd1045
·
1 Parent(s): 0d908d7

fixing ensure img greyscaled

Browse files
Files changed (2) hide show
  1. app.py +1 -0
  2. modelbuilder.py +8 -5
app.py CHANGED
@@ -43,6 +43,7 @@ for example in dataset:
43
  # 3️⃣ Define preprocessing and inference function
44
  # ------------------------------------------------------------
45
  def preprocess_image(img: Image.Image):
 
46
  img = img.resize(TARGET_SIZE, Image.BILINEAR) # resize image to (target_height, target_width)
47
  img_array = img_to_array(img) / 255.0 # convert image to float32 NumPy array and normalize pixel values to [0, 1]
48
  img_array = np.expand_dims(img_array, axis=0) # add batch dimension -> shape becomes (1, height, width, channels)
 
43
  # 3️⃣ Define preprocessing and inference function
44
  # ------------------------------------------------------------
45
  def preprocess_image(img: Image.Image):
46
+ img = img.convert("L") # Ensure grayscale (1 channel): "L" = 8-bit grayscale
47
  img = img.resize(TARGET_SIZE, Image.BILINEAR) # resize image to (target_height, target_width)
48
  img_array = img_to_array(img) / 255.0 # convert image to float32 NumPy array and normalize pixel values to [0, 1]
49
  img_array = np.expand_dims(img_array, axis=0) # add batch dimension -> shape becomes (1, height, width, channels)
modelbuilder.py CHANGED
@@ -61,7 +61,7 @@ class ModelBuilder:
61
  )
62
 
63
  def get_augmentation_pipe(self):
64
- # Random-* layers are stochastic only when training=True
65
  # disabled during inference/evaluation
66
  return Sequential(
67
  [
@@ -78,12 +78,15 @@ class ModelBuilder:
78
 
79
  # Define input layer
80
  inputs = Input(shape=self.input_shape, name="inputs")
81
- # Random-* layers are stochastic only when training=True
82
-
83
  x_aug = self.get_augmentation_pipe()(
84
  inputs
85
- ) # stochastic only when training=True
86
- x = Rescaling(1.0 / 255)(x_aug) # disabled during inference/evaluation
 
 
 
87
 
88
  # Model selector
89
  match self.model_type:
 
61
  )
62
 
63
  def get_augmentation_pipe(self):
64
+ # Random/Augmentation layers are stochastic only when training=True
65
  # disabled during inference/evaluation
66
  return Sequential(
67
  [
 
78
 
79
  # Define input layer
80
  inputs = Input(shape=self.input_shape, name="inputs")
81
+
82
+ # --- Random/Augmentation layers are stochastic only when training=True
83
  x_aug = self.get_augmentation_pipe()(
84
  inputs
85
+ )
86
+ # ----- end augmentation -----
87
+
88
+ # --- common preprocessing layer: rescaling to [0,1]
89
+ x = Rescaling(1.0 / 255)(x_aug)
90
 
91
  # Model selector
92
  match self.model_type: