improve the speed of face recognition on CPU

[arad2022]

How I can increase the speed of below program I can use GPU in this time I want to increase speed on CPU. Any suggestion? Why all the model loaded and how I can disable them?

Import libraries

import torch import insightface import cv2 import numpy as np

Load a face recognition model

model = insightface.app.FaceAnalysis() model.prepare(ctx_id = -1) # use CPU and set NMS threshold #model.prepare(ctx_id = 0)

Load two images

img1 = cv2.imread(r"D:\20240127_120414.png")

img2 = cv2.imread(r"D:\code\IMG_20240205_140538.jpg")

faces1 = model.get(img1)

box = faces1[0]['bbox'] bbox = box.astype(int)

x_min, y_min, x_max, y_max = int(box['left']), int(box['top']), int(box['right']), int(box['bottom'])

face_img = img1[bbox[1]:bbox[3], bbox[0]:bbox[2]] print(face_img.shape) cv2.imwrite("./test.png", face_img) faces2 = model.get(img2)

Extract face embeddings

embeddings1 = np.array([face.normed_embedding for face in faces1]) embeddings2 = np.array([face.normed_embedding for face in faces2])

Compute cosine similarity between embeddings

similarity = np.dot(embeddings1, embeddings2.T)

output Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: C:\Users\windows/.insightface\models\buffalo_l\1k3d68.onnx landmark_3d_68 ['None', 3, 192, 192] 0.0 1.0 Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: C:\Users\windows/.insightface\models\buffalo_l\2d106det.onnx landmark_2d_106 ['None', 3, 192, 192] 0.0 1.0 Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: C:\Users\windows/.insightface\models\buffalo_l\det_10g.onnx detection [1, 3, '?', '?'] 127.5 128.0 Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: C:\Users\windows/.insightface\models\buffalo_l\genderage.onnx genderage ['None', 3, 96, 96] 0.0 1.0 Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: C:\Users\windows/.insightface\models\buffalo_l\w600k_r50.onnx recognition ['None', 3, 112, 112] 127.5 127.5 set det-size: (640, 640)

[longduong2]

Using OpenVino to export this module lets you rebuild the Face Analysis flow for CPU, making it run much faster.

[arad2022]

Using OpenVino to export this module lets you rebuild the Face Analysis flow for CPU, making it run much faster.

Could you please provide code example for that? How can I export Buffalo model to OpenVino format and then load it for test?

[Raghucharan16]

I have downloaded the glint360k model and how can i use that for verification when i extracted it these are the files inside it.