API has convert("BGR") when loading a jpg image?

[AmrRahmy]

i am more familiar with java and c# but while debugging an error in the api endpoint,
/represent or http://localhost:5000/represent

it seems that if you supply a regular image, it errors out with,

{
    "error": "Exception while representing: conversion from RGB to BGR not supported"
}

looking into it, it seems the file 'preprocessing.py', function 'def load_image(img: Union[str, np.ndarray]) -> Tuple[np.ndarray, str]:'

if img.startswith("http"):
        return (
            np.array(Image.open(requests.get(img, stream=True, timeout=60).raw).convert("BGR")),
            # return url as image name
            img,
        )

instead of

if img.startswith("http"):
        return (
            np.array(Image.open(requests.get(img, stream=True, timeout=60).raw)),
            # return url as image name
            img,
        )

as if it's expecting an opencv compatible image or mat? even if the detector_backend is set to 'retinaface'

http request body tried,

{
    "model_name": "Facenet",
    "img": "http://192.168.3.90/sharedDirectory/face.jpg",
    "detector_backend": "retinaface",
    "distance_metric": "cosine"
}

{
    "img": "http://192.168.3.90/sharedDirectory/face.jpg"
}

url is valid and accessible.

[serengil]

Would you please share the image? Could it be gray scale?

[AmrRahmy]

it's from celebrity faces dataset

[serengil]

Oh I see the root cause of the bug. Will fix this today.

[serengil]

Closed with PR - #1026