Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -71,10 +71,11 @@ except Exception as e:
|
|
71 |
print(f"Error loading model: {str(e)}")
|
72 |
traceback.print_exc()
|
73 |
|
|
|
74 |
transform = transforms.Compose([
|
75 |
transforms.Resize((128, 128)),
|
76 |
-
transforms.
|
77 |
-
transforms.
|
78 |
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
|
79 |
])
|
80 |
|
@@ -83,17 +84,14 @@ def process_image(image):
|
|
83 |
return None
|
84 |
|
85 |
try:
|
86 |
-
# numpy array
|
87 |
if isinstance(image, np.ndarray):
|
88 |
image = Image.fromarray(image.astype('uint8'))
|
89 |
|
90 |
-
#
|
91 |
if image.mode != 'RGB':
|
92 |
image = image.convert('RGB')
|
93 |
|
94 |
-
# 이미지 크기 조정
|
95 |
-
image = image.resize((128, 128), Image.Resampling.LANCZOS)
|
96 |
-
|
97 |
print(f"Processed image size: {image.size}")
|
98 |
print(f"Processed image mode: {image.mode}")
|
99 |
|
@@ -108,32 +106,21 @@ def predict(image):
|
|
108 |
return {cls: 0.0 for cls in ["Rope", "Hammer", "Other"]}
|
109 |
|
110 |
try:
|
111 |
-
#
|
112 |
processed_image = process_image(image)
|
113 |
if processed_image is None:
|
114 |
return {cls: 0.0 for cls in ["Rope", "Hammer", "Other"]}
|
115 |
|
116 |
-
#
|
117 |
try:
|
118 |
-
|
119 |
-
img_array = np.array(processed_image)
|
120 |
-
# numpy array를 torch tensor로 변환
|
121 |
-
tensor_image = torch.from_numpy(img_array.transpose((2, 0, 1))).float() / 255.0
|
122 |
-
# 정규화
|
123 |
-
tensor_image = transforms.Normalize(
|
124 |
-
mean=[0.485, 0.456, 0.406],
|
125 |
-
std=[0.229, 0.224, 0.225]
|
126 |
-
)(tensor_image)
|
127 |
-
# 배치 차원 추가
|
128 |
-
tensor_image = tensor_image.unsqueeze(0)
|
129 |
-
|
130 |
print(f"Input tensor shape: {tensor_image.shape}")
|
131 |
except Exception as e:
|
132 |
print(f"Error in tensor conversion: {str(e)}")
|
133 |
traceback.print_exc()
|
134 |
return {cls: 0.0 for cls in ["Rope", "Hammer", "Other"]}
|
135 |
|
136 |
-
#
|
137 |
with torch.no_grad():
|
138 |
outputs = model(tensor_image)
|
139 |
print(f"Raw outputs: {outputs}")
|
@@ -141,7 +128,7 @@ def predict(image):
|
|
141 |
probabilities = F.softmax(outputs, dim=1)[0].cpu().numpy()
|
142 |
print(f"Probabilities: {probabilities}")
|
143 |
|
144 |
-
#
|
145 |
classes = ["Rope", "Hammer", "Other"]
|
146 |
results = {cls: float(prob) for cls, prob in zip(classes, probabilities)}
|
147 |
print(f"Final results: {results}")
|
@@ -152,7 +139,7 @@ def predict(image):
|
|
152 |
traceback.print_exc()
|
153 |
return {cls: 0.0 for cls in ["Rope", "Hammer", "Other"]}
|
154 |
|
155 |
-
# Gradio
|
156 |
interface = gr.Interface(
|
157 |
fn=predict,
|
158 |
inputs=gr.Image(),
|
@@ -161,6 +148,6 @@ interface = gr.Interface(
|
|
161 |
description="Upload an image of a tool to classify it as 'Rope', 'Hammer', or 'Other'.",
|
162 |
)
|
163 |
|
164 |
-
#
|
165 |
if __name__ == "__main__":
|
166 |
interface.launch()
|
|
|
71 |
print(f"Error loading model: {str(e)}")
|
72 |
traceback.print_exc()
|
73 |
|
74 |
+
# Define image transformation pipeline
|
75 |
transform = transforms.Compose([
|
76 |
transforms.Resize((128, 128)),
|
77 |
+
transforms.PILToTensor(), # Changed from ToTensor()
|
78 |
+
transforms.ConvertImageDtype(torch.float32),
|
79 |
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
|
80 |
])
|
81 |
|
|
|
84 |
return None
|
85 |
|
86 |
try:
|
87 |
+
# Convert numpy array to PIL Image
|
88 |
if isinstance(image, np.ndarray):
|
89 |
image = Image.fromarray(image.astype('uint8'))
|
90 |
|
91 |
+
# Convert to RGB if necessary
|
92 |
if image.mode != 'RGB':
|
93 |
image = image.convert('RGB')
|
94 |
|
|
|
|
|
|
|
95 |
print(f"Processed image size: {image.size}")
|
96 |
print(f"Processed image mode: {image.mode}")
|
97 |
|
|
|
106 |
return {cls: 0.0 for cls in ["Rope", "Hammer", "Other"]}
|
107 |
|
108 |
try:
|
109 |
+
# Process the image
|
110 |
processed_image = process_image(image)
|
111 |
if processed_image is None:
|
112 |
return {cls: 0.0 for cls in ["Rope", "Hammer", "Other"]}
|
113 |
|
114 |
+
# Transform image to tensor using torchvision transforms
|
115 |
try:
|
116 |
+
tensor_image = transform(processed_image).unsqueeze(0)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
117 |
print(f"Input tensor shape: {tensor_image.shape}")
|
118 |
except Exception as e:
|
119 |
print(f"Error in tensor conversion: {str(e)}")
|
120 |
traceback.print_exc()
|
121 |
return {cls: 0.0 for cls in ["Rope", "Hammer", "Other"]}
|
122 |
|
123 |
+
# Make prediction
|
124 |
with torch.no_grad():
|
125 |
outputs = model(tensor_image)
|
126 |
print(f"Raw outputs: {outputs}")
|
|
|
128 |
probabilities = F.softmax(outputs, dim=1)[0].cpu().numpy()
|
129 |
print(f"Probabilities: {probabilities}")
|
130 |
|
131 |
+
# Return results
|
132 |
classes = ["Rope", "Hammer", "Other"]
|
133 |
results = {cls: float(prob) for cls, prob in zip(classes, probabilities)}
|
134 |
print(f"Final results: {results}")
|
|
|
139 |
traceback.print_exc()
|
140 |
return {cls: 0.0 for cls in ["Rope", "Hammer", "Other"]}
|
141 |
|
142 |
+
# Gradio interface
|
143 |
interface = gr.Interface(
|
144 |
fn=predict,
|
145 |
inputs=gr.Image(),
|
|
|
148 |
description="Upload an image of a tool to classify it as 'Rope', 'Hammer', or 'Other'.",
|
149 |
)
|
150 |
|
151 |
+
# Launch the interface
|
152 |
if __name__ == "__main__":
|
153 |
interface.launch()
|