Prediction wrong y_min coordinates when calling /vision/face/recognize

There might be a potential bug in the face recognition API in the latest Deepstack version (2021.02.1).

When calling /vision/face/recognize on an image with more than one face, I get the same (first predicted face) y_min value.

Request

curl --location --request POST 'http://[MY_DEEPSTACK_SERVER]/v1/vision/face/recognize' --form 'image=@"/idriselba2.jpg"' --form 'min_confidence="0"'

Response

{
    "success": true,
    "predictions": [
        {
            "confidence": 0,
            "userid": "unknown",
            "y_min": 226,
            "x_min": 1627,
            "y_max": 693,
            "x_max": 1991
        },
        {
            "confidence": 0,
            "userid": "unknown",
            "y_min": 226,
            "x_min": 869,
            "y_max": 717,
            "x_max": 1198
        }
    ],
    "duration": 0
}

Log output

DeepStack: Version 2021.02.1
/v1/vision/face
---------------------------------------
/v1/vision/face/recognize
---------------------------------------
/v1/vision/face/register
---------------------------------------
/v1/vision/face/match
---------------------------------------
/v1/vision/face/list
---------------------------------------
/v1/vision/face/delete
---------------------------------------
/v1/vision/detection
---------------------------------------
---------------------------------------
v1/backup
---------------------------------------
v1/restore
[GIN] 2021/05/04 - 17:40:07 | 200 |  3.511393478s |     192.168.0.1 | POST     /v1/vision/face/recognize

Same bug also reported here: https://github.com/iceglow/node-red-contrib-deepstack/issues/6