how to use gpu limit

[mickey-kim]

hello

i use face_recognition and i have some question
can i limit gpu usage??

[alessiosavi]

The problem is related to the high dimension of the image.

A simple and basic approach, can be resize the photo.
Instead of use the builtin method load_image_file, you can use mine:
https://github.com/alessiosavi/PyRecognizer/blob/50b9fe1e4676cb85aa9f88c15fbe988a258eec1d/utils/util.py#L286

It will simply resize the photo. You can tweak for you purpouse.

def load_image_file(file, mode='RGB',):
    """
    Loads an image file (.jpg, .png, etc) into a numpy array
    :param file: image file name or file object to load
    :param mode: format to convert the image to. Only 'RGB' (8-bit RGB, 3 channels) and 'L' (black and white) are supported.
    :return: image contents as numpy array
    """

    im = PIL.Image.open(file)
    width, height = im.size
    w, h = width, height
    log = logging.getLogger()

    ratio = -1
    # Ratio for resize the image
    log.debug("load_image_file | Image dimension: ({}:{})".format(w, h))
    # Resize in case of to bigger dimension
    if 1200 <= width <= 1600 or 1200 <= height <= 1600:
        ratio = 1/2
    elif 1600 <= width <= 3600 or 1600 <= height <= 3600:
        ratio = 1/3
    elif width > 3600 or height > 3600:
        if width > height:
            ratio = width/800
        else:
            ratio = height/800
        log.debug("Dimension: w: {} | h: {}".format(w, h))
        log.debug("new ratio -> {}".format(ratio))

    if 0 < ratio < 1:
        # Scale image in case of width > 1600
        w = width * ratio
        h = height * ratio
    elif ratio > 1:
        # Scale image in case of width > 3600
        w = width / ratio
        h = height / ratio
    if w != width:
        # Check if scaling was applied
        maxsize = (w, h)
        log.debug(
            "Image have to high dimension, avoiding memory error. Resizing to {}".format(maxsize))
        im.thumbnail(maxsize, PIL.Image.ANTIALIAS)

    if mode:
        im = im.convert(mode)
    return np.array(im), ratio