casting to float32
Browse files
app.py
CHANGED
@@ -79,10 +79,10 @@ def get_activations(intermediate_model, image: list,
|
|
79 |
|
80 |
def predict_and_analyze(model_name, num_channels, dim, image):
|
81 |
|
82 |
-
'''
|
|
|
83 |
|
84 |
The image must be a numpy array of shape (C, W, W) or (1, C, W, W)
|
85 |
-
|
86 |
'''
|
87 |
|
88 |
num_channels = int(num_channels)
|
@@ -90,6 +90,7 @@ def predict_and_analyze(model_name, num_channels, dim, image):
|
|
90 |
|
91 |
print("Loading data")
|
92 |
image = np.load(image.name, allow_pickle=True)
|
|
|
93 |
|
94 |
if len(image.shape) != 4:
|
95 |
image = image[np.newaxis, :, :, :]
|
@@ -97,7 +98,7 @@ def predict_and_analyze(model_name, num_channels, dim, image):
|
|
97 |
assert image.shape == (1, num_channels, W, W), "Data is the wrong shape"
|
98 |
|
99 |
model_name += '_%i' % (num_channels)
|
100 |
-
|
101 |
print("Loading model")
|
102 |
model = load_model(model_name, activation=True)
|
103 |
print("Model loaded")
|
@@ -153,7 +154,7 @@ if __name__ == "__main__":
|
|
153 |
|
154 |
demo = gr.Interface(
|
155 |
fn=predict_and_analyze,
|
156 |
-
inputs=[gr.Dropdown(["
|
157 |
value="efficientnet",
|
158 |
label="Model Selection",
|
159 |
show_label=True),
|
|
|
79 |
|
80 |
def predict_and_analyze(model_name, num_channels, dim, image):
|
81 |
|
82 |
+
'''
|
83 |
+
Loads a model with activations, passes through image and shows activations
|
84 |
|
85 |
The image must be a numpy array of shape (C, W, W) or (1, C, W, W)
|
|
|
86 |
'''
|
87 |
|
88 |
num_channels = int(num_channels)
|
|
|
90 |
|
91 |
print("Loading data")
|
92 |
image = np.load(image.name, allow_pickle=True)
|
93 |
+
image = image.astype(np.float32)
|
94 |
|
95 |
if len(image.shape) != 4:
|
96 |
image = image[np.newaxis, :, :, :]
|
|
|
98 |
assert image.shape == (1, num_channels, W, W), "Data is the wrong shape"
|
99 |
|
100 |
model_name += '_%i' % (num_channels)
|
101 |
+
|
102 |
print("Loading model")
|
103 |
model = load_model(model_name, activation=True)
|
104 |
print("Model loaded")
|
|
|
154 |
|
155 |
demo = gr.Interface(
|
156 |
fn=predict_and_analyze,
|
157 |
+
inputs=[gr.Dropdown(["efficientnet", "regnet"],
|
158 |
value="efficientnet",
|
159 |
label="Model Selection",
|
160 |
show_label=True),
|