DeepFace’s version
0.0.91

Python version
3.11.1

Operating System
Windows 10

Need guidance on how to increase the accuracy, this is the code I’m using

backends = [
    'opencv',
    'ssd',
    'dlib',
    'mtcnn',
    'retinaface',
    'mediapipe',
    'yolov8',
    'yunet',
    'fastmtcnn',
]


def find_the_face(filename_path, user_dir):
    dfs = DeepFace.find(
        img_path=filename_path,
        db_path=f"{user_dir}",
        enforce_detection=True,
        detector_backend=backends[1]
    )

    results = []
    for df in dfs:
        df_json = json.loads(df.to_json(orient='records'))
        results += df_json

    return results

By example:
If I use this image of Kate

Then I get these results:

{
"found_identities": {
"ana_de_armas": 3,
"anne_hathaway": 2,
"cameron_diaz": 1,
"chris_evans": 1,
"daniel_radcliffe": 2,
"edgar_ramirez": 1,
"eiza_gonzalez": 3,
"emilia_clarke": 3,
"emily_blunt": 5,
"emma_stone": 1,
"henry_cavil": 3,
"jennifer_lawrence": 1,
"kate_winlest": 6,
"lily_james": 2,
"margot_robbie": 1,
"mario_casas": 1,
"melanie_laurent": 2,
"mila_kunis": 2,
"robert_de_niro": 2,
"ryan_reynolds": 2,
"sam_claflin": 1,
"scarlett_johansson": 1,
"tom_hardy": 1,
"zazie_beetz": 1,
"zendaya": 1
},
"msg": "FACES_FOUND"
}

But if I use this image of Kate

Then I get these results

{
"found_identities": {
"ana_de_armas": 1,
"emily_blunt": 5,
"kate_winlest": 5
},
"msg": "FACES_FOUND"
}

Notice my personal Emily’s database