Update pipeline.py
Browse files- pipeline.py +19 -21
pipeline.py
CHANGED
|
@@ -46,36 +46,34 @@ class PreTrainedPipeline(Pipeline):
|
|
| 46 |
# Resize image to expected size
|
| 47 |
|
| 48 |
expected_input_size = self.model.input_shape
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
if expected_input_size[-1] == 1:
|
| 50 |
inputs = inputs.convert("L")
|
| 51 |
|
|
|
|
|
|
|
| 52 |
target_size = (expected_input_size[1], expected_input_size[2])
|
| 53 |
img = tf.image.resize(inputs, target_size)
|
| 54 |
-
|
| 55 |
img_array = tf.keras.preprocessing.image.img_to_array(img)
|
| 56 |
img_array = img_array[tf.newaxis, ...]
|
| 57 |
-
|
| 58 |
-
predictions = self.model.predict(img_array
|
| 59 |
|
| 60 |
self.single_output_unit = (
|
| 61 |
self.model.output_shape[1] == 1
|
| 62 |
) # if there are two classes
|
| 63 |
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
labels
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
{
|
| 75 |
-
"label": str(self.id2label[str(i)]),
|
| 76 |
-
"mask": base64.b64encode(predictions[0][i]),
|
| 77 |
-
"score": float(score),
|
| 78 |
-
}
|
| 79 |
-
for i, score in enumerate(predictions[0])
|
| 80 |
-
]
|
| 81 |
-
return sorted(labels, key=lambda tup: tup["score"], reverse=True)[: self.top_k]
|
|
|
|
| 46 |
# Resize image to expected size
|
| 47 |
|
| 48 |
expected_input_size = self.model.input_shape
|
| 49 |
+
|
| 50 |
+
|
| 51 |
+
with Image.open(inputs) as im:
|
| 52 |
+
inputs = np.array(im)
|
| 53 |
if expected_input_size[-1] == 1:
|
| 54 |
inputs = inputs.convert("L")
|
| 55 |
|
| 56 |
+
|
| 57 |
+
|
| 58 |
target_size = (expected_input_size[1], expected_input_size[2])
|
| 59 |
img = tf.image.resize(inputs, target_size)
|
| 60 |
+
|
| 61 |
img_array = tf.keras.preprocessing.image.img_to_array(img)
|
| 62 |
img_array = img_array[tf.newaxis, ...]
|
| 63 |
+
|
| 64 |
+
predictions = self.model.predict(img_array)
|
| 65 |
|
| 66 |
self.single_output_unit = (
|
| 67 |
self.model.output_shape[1] == 1
|
| 68 |
) # if there are two classes
|
| 69 |
|
| 70 |
+
labels = []
|
| 71 |
+
for i in enumerate(predictions):
|
| 72 |
+
|
| 73 |
+
labels.append({
|
| 74 |
+
"label": str(i[0]),
|
| 75 |
+
"mask": base64.b64encode(i[1]),
|
| 76 |
+
"score": 1.0,
|
| 77 |
+
})
|
| 78 |
+
|
| 79 |
+
return sorted(labels)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|