deepface: Don't print logs from MTCNN backend

Question

I have a very simple code that I use for face detection from an image, for example:

from deepface.commons import functions
import numpy as np

random_image = np.random.randint(
    0, 255, size=(360, 360, 3)
)

detected_face = functions.detect_face(
    img=random_image,
    detector_backend="mtcnn",
    enforce_detection=False,
)[0]

This code prints out the following logs (made by MTCNN backend):

1/1 [==============================] - 0s 24ms/step
1/1 [==============================] - 0s 15ms/step
1/1 [==============================] - 0s 13ms/step
1/1 [==============================] - 0s 11ms/step
1/1 [==============================] - 0s 11ms/step
1/1 [==============================] - 0s 11ms/step
1/1 [==============================] - 0s 10ms/step
1/1 [==============================] - 0s 9ms/step
1/1 [==============================] - 0s 10ms/step
5/5 [==============================] - 0s 4ms/step
1/1 [==============================] - 0s 16ms/step

Is there a way how to suppress deepface to print the logs, please?

score 2 · Answer 1 · answered Jan 29 '23 at 11:08

Somtimes when you cannot supress printing the messages, you can decoreate function and capture the output.

from functools import wraps
from deepface.commons import functions
import sys
import io


def capture_output(func):
    """Wrapper to capture print output."""

    @wraps(func)
    def wrapper(*args, **kwargs):
        old_stdout = sys.stdout
        new_stdout = io.StringIO()
        sys.stdout = new_stdout
        try:
            return func(*args, **kwargs)
        finally:
            sys.stdout = old_stdout

    return wrapper

w_detect_face = capture_output(functions.detect_face)

detected_face = w_detect_face(
    img=random_image,
    detector_backend="mtcnn",
    enforce_detection=False,
)[0]

This answer should be marked as accepted, it works for any functions. — let me down slowly, Aug 02 '23 at 09:46

score 2 · Answer 2 · answered May 15 '23 at 22:07

please reach to mtcnn.py file in your pc. It is perhaps in the path ~Python\Python311\Lib\site-packages\ in windows OS. add verbose=0 in lines: 342, 410 and 466:

     line 342: out = self._pnet.predict(img_y)
     Line 410: out = self._rnet.predict(tempimg1) 
     Line 466: out = self._onet.predict(tempimg1)

like this:

     line 342: out = self._pnet.predict(img_y,verbose=0)
     Line 410: out = self._rnet.predict(tempimg1,verbose=0) 
     Line 466: out = self._onet.predict(tempimg1,verbose=0)

`

This actually works and seems to be the only working solution. The only problem is it's not ideal as you'll have to patch the dependent libs of deepface. — nonsequiter, Jul 16 '23 at 12:49

deepface: Don't print logs from MTCNN backend

2 Answers2