Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
Update app.py
Browse files
app.py
CHANGED
|
@@ -7,23 +7,41 @@ import torch
|
|
| 7 |
import numpy as np
|
| 8 |
|
| 9 |
def preprocess_image(img):
|
|
|
|
|
|
|
|
|
|
| 10 |
# Convert numpy array to Tensor and ensure correct shape
|
| 11 |
if isinstance(img, np.ndarray):
|
| 12 |
img = K.image_to_tensor(img, keepdim=False).float() / 255.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
|
| 14 |
# Ensure 3D tensor (C, H, W)
|
| 15 |
if img.ndim == 2:
|
| 16 |
img = img.unsqueeze(0)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
# Ensure 3 channel image
|
| 19 |
if img.shape[0] == 1:
|
| 20 |
-
img = img.
|
| 21 |
elif img.shape[0] > 3:
|
| 22 |
img = img[:3] # Take only the first 3 channels if more than 3
|
| 23 |
|
|
|
|
|
|
|
| 24 |
# Add batch dimension
|
| 25 |
img = img.unsqueeze(0)
|
| 26 |
|
|
|
|
| 27 |
return img
|
| 28 |
|
| 29 |
def inference(img_1, img_2):
|
|
|
|
| 7 |
import numpy as np
|
| 8 |
|
| 9 |
def preprocess_image(img):
|
| 10 |
+
print(f"Input image type: {type(img)}")
|
| 11 |
+
print(f"Input image shape: {img.shape if hasattr(img, 'shape') else 'No shape attribute'}")
|
| 12 |
+
|
| 13 |
# Convert numpy array to Tensor and ensure correct shape
|
| 14 |
if isinstance(img, np.ndarray):
|
| 15 |
img = K.image_to_tensor(img, keepdim=False).float() / 255.0
|
| 16 |
+
elif isinstance(img, torch.Tensor):
|
| 17 |
+
img = img.float()
|
| 18 |
+
if img.max() > 1.0:
|
| 19 |
+
img = img / 255.0
|
| 20 |
+
else:
|
| 21 |
+
raise ValueError(f"Unsupported image type: {type(img)}")
|
| 22 |
+
|
| 23 |
+
print(f"After conversion to tensor - shape: {img.shape}")
|
| 24 |
|
| 25 |
# Ensure 3D tensor (C, H, W)
|
| 26 |
if img.ndim == 2:
|
| 27 |
img = img.unsqueeze(0)
|
| 28 |
+
elif img.ndim == 3 and img.shape[0] not in [1, 3]:
|
| 29 |
+
img = img.permute(2, 0, 1)
|
| 30 |
+
|
| 31 |
+
print(f"After ensuring 3D - shape: {img.shape}")
|
| 32 |
|
| 33 |
# Ensure 3 channel image
|
| 34 |
if img.shape[0] == 1:
|
| 35 |
+
img = img.expand(3, -1, -1)
|
| 36 |
elif img.shape[0] > 3:
|
| 37 |
img = img[:3] # Take only the first 3 channels if more than 3
|
| 38 |
|
| 39 |
+
print(f"After ensuring 3 channels - shape: {img.shape}")
|
| 40 |
+
|
| 41 |
# Add batch dimension
|
| 42 |
img = img.unsqueeze(0)
|
| 43 |
|
| 44 |
+
print(f"Final tensor shape: {img.shape}")
|
| 45 |
return img
|
| 46 |
|
| 47 |
def inference(img_1, img_2):
|