Spaces:
Runtime error
Runtime error
supersolar
commited on
Update infer.py
Browse files
infer.py
CHANGED
@@ -53,13 +53,14 @@ def infer_pipe(pipe, image_input, task_name, seed, device):
|
|
53 |
).images[0]
|
54 |
|
55 |
# Post-process the prediction
|
|
|
56 |
if task_name == 'depth':
|
57 |
output_npy = pred.mean(axis=-1)
|
58 |
-
|
|
|
59 |
else:
|
60 |
output_npy = pred
|
61 |
output_color = Image.fromarray((output_npy * 255).astype(np.uint8))
|
62 |
-
|
63 |
return output_color
|
64 |
|
65 |
def lotus_video(input_video, task_name, seed, device):
|
@@ -121,14 +122,15 @@ def lotus_video(input_video, task_name, seed, device):
|
|
121 |
task_emb=task_emb,
|
122 |
).images[0]
|
123 |
# Post-process the prediction
|
|
|
124 |
if task_name == 'depth':
|
125 |
output_npy_g = pred_g.mean(axis=-1)
|
126 |
-
|
|
|
127 |
else:
|
128 |
output_npy_g = pred_g
|
129 |
output_color_g = Image.fromarray((output_npy_g * 255).astype(np.uint8))
|
130 |
-
|
131 |
-
output_g.append(output_color_g)
|
132 |
|
133 |
|
134 |
return output_g
|
@@ -305,13 +307,14 @@ def main():
|
|
305 |
|
306 |
# Post-process the prediction
|
307 |
save_file_name = os.path.basename(test_images[i])[:-4]
|
308 |
-
|
|
|
309 |
output_npy = pred.mean(axis=-1)
|
310 |
-
|
|
|
311 |
else:
|
312 |
output_npy = pred
|
313 |
output_color = Image.fromarray((output_npy * 255).astype(np.uint8))
|
314 |
-
|
315 |
output_color.save(os.path.join(output_dir_color, f'{save_file_name}.png'))
|
316 |
np.save(os.path.join(output_dir_npy, f'{save_file_name}.npy'), output_npy)
|
317 |
|
|
|
53 |
).images[0]
|
54 |
|
55 |
# Post-process the prediction
|
56 |
+
# 在 infer_pipe 函数中
|
57 |
if task_name == 'depth':
|
58 |
output_npy = pred.mean(axis=-1)
|
59 |
+
# 修改为输出灰度图
|
60 |
+
output_color = Image.fromarray((output_npy * 255).astype(np.uint8), mode='L')
|
61 |
else:
|
62 |
output_npy = pred
|
63 |
output_color = Image.fromarray((output_npy * 255).astype(np.uint8))
|
|
|
64 |
return output_color
|
65 |
|
66 |
def lotus_video(input_video, task_name, seed, device):
|
|
|
122 |
task_emb=task_emb,
|
123 |
).images[0]
|
124 |
# Post-process the prediction
|
125 |
+
# 在 lotus_video 函数中
|
126 |
if task_name == 'depth':
|
127 |
output_npy_g = pred_g.mean(axis=-1)
|
128 |
+
# 修改为输出灰度图
|
129 |
+
output_color_g = Image.fromarray((output_npy_g * 255).astype(np.uint8), mode='L')
|
130 |
else:
|
131 |
output_npy_g = pred_g
|
132 |
output_color_g = Image.fromarray((output_npy_g * 255).astype(np.uint8))
|
133 |
+
output_g.append(output_color_g)
|
|
|
134 |
|
135 |
|
136 |
return output_g
|
|
|
307 |
|
308 |
# Post-process the prediction
|
309 |
save_file_name = os.path.basename(test_images[i])[:-4]
|
310 |
+
# 在 infer_pipe 函数中
|
311 |
+
if task_name == 'depth':
|
312 |
output_npy = pred.mean(axis=-1)
|
313 |
+
# 修改为输出灰度图
|
314 |
+
output_color = Image.fromarray((output_npy * 255).astype(np.uint8), mode='L')
|
315 |
else:
|
316 |
output_npy = pred
|
317 |
output_color = Image.fromarray((output_npy * 255).astype(np.uint8))
|
|
|
318 |
output_color.save(os.path.join(output_dir_color, f'{save_file_name}.png'))
|
319 |
np.save(os.path.join(output_dir_npy, f'{save_file_name}.npy'), output_npy)
|
320 |
|