How to save and load face feature database to disk in python?

[nauyeskhan-monosoft]

I am using InspireFace’s FeatureHub in Python to handle face recognition features, and I would like to save the extracted features for millions of images in a persistent database file. My goal is to avoid reprocessing and reinserting each feature on every run.

Currently, I’ve configured FeatureHub as follows:

feature_hub_config = ifac.FeatureHubConfiguration(
    feature_block_num=20,
    enable_use_db=True,
    db_path="face_features.db",
    search_threshold=-1,
    search_mode=ifac.HF_SEARCH_MODE_EXHAUSTIVE,
)
ret = ifac.feature_hub_enable(feature_hub_config)
assert ret, "Failed to enable FeatureHub."

for idx, image_name in tqdm(enumerate(os.listdir(origin_database_path))):
    image_path = os.path.join(origin_database_path, image_name)
    if image_path.endswith((".jpg", ".jpeg", ".png")):
        image = cv2.imread(image_path)
        assert image is not None, f"Failed to load image {image_path}"
        faces = session.face_detection(image)
        if faces:
            face = faces[0]  # Use the first detected face
            feature = session.face_feature_extract(image, face)
            identity = ifac.FaceIdentity(feature, custom_id=idx, tag=image_name)
            ret = ifac.feature_hub_face_insert(identity)
            assert ret, f"Failed to insert face for {image_name}"

Despite setting enable_use_db=True and specifying db_path, I don’t see any database file saved at the specified location. Is there an option or additional method to save the FeatureHub to disk, so that it can be loaded directly in a new session without re-insertion? If so, could you provide guidance on how to do this in Python?

[tunmx]

Thanks for your feedback. Theoretically, if you specify db_path, it should be stored. If not, there may be some problems, I will test the latest version later and give you feedback.

[nauyeskhan-monosoft]

Hello!

@tunmx, any updates?

[tunmx]

Sorry, I haven't updated it yet, maybe it needs to be on the weekend.

[tunmx]

Hi, I've fixed this issue and added Python test cases. It occurred in the Python native interface, so you just need to pull the code - no need to recompile the dynamic library. #112

import os
import cv2
import inspireface as ifac
from inspireface.param import *
import numpy as np
import os

def case_feature_hub():
    db_path = "test.db"
    if os.path.exists(db_path):
        os.remove(db_path)
    # Configure the feature management system.
    feature_hub_config = ifac.FeatureHubConfiguration(
        feature_block_num=10,
        enable_use_db=True,
        db_path=db_path,
        search_threshold=0.48,
        search_mode=HF_SEARCH_MODE_EAGER,
    )
    ret = ifac.feature_hub_enable(feature_hub_config)
    assert ret, "Failed to enable FeatureHub."
    print(ifac.feature_hub_get_face_count())
    for i in range(10):
        feature = ifac.FaceIdentity(np.random.rand(512), i, "test")
        ifac.feature_hub_face_insert(feature)
    print(ifac.feature_hub_get_face_count())

    assert os.path.exists(db_path), "FeatureHub database file not found."

if __name__ == "__main__":
    case_feature_hub()