隐藏终端的 ultralytics model.predict() 输出

Question

我有这个由

model.predict()

生成的输出

0: 480x640 1 Hole, 234.1ms
Speed: 3.0ms preprocess, 234.1ms inference, 4.0ms postprocess per image at shape (1, 3, 640, 640)

0: 480x640 1 Hole, 193.6ms
Speed: 3.0ms preprocess, 193.6ms inference, 3.5ms postprocess per image at shape (1, 3, 640, 640)

...

如何隐藏终端的输出？

我在这个官方链接里查不到信息 https://docs.ultralytics.com/modes/predict/#arguments

Answer 1

实现此目的的一种方法是在 os 模块和 contextlib 模块的帮助下将标准输出重定向到 dev/null （实际上会丢弃所有输出，您甚至看不到错误）：

import os
from contextlib import redirect_stdout

with open(os.devnull, 'w') as devnull:
  with redirect_stdout(devnull):
    result = modle.predict

Answer 2

Ultralytics 文档很遗憾不是最新的。正确的做法是：

from ultralytics import YOLO

model = YOLO('yolov8m-seg.pt')
results = model.predict(source='0', verbose=False)
for result in results:
    masks = result.masks.masks
    print(masks.shape)

注意

verbose=False

论点。这不会打印默认输出

...
0: 480x640 1 Hole, 193.6ms
Speed: 3.0ms preprocess, 193.6ms inference, 3.5ms postprocess per image at shape (1, 3, 640, 640)
...

仅在这种情况下：

...
torch.Size([4, 480, 640])
...