FaceImageKit

Python实现的人脸图像相关算法Pipeline

Install

python

example:

uv venv --python 3.12

deploy

cpu

uv pip install -e .[cpu]

cuda

uv pip install -e .[cuda]

develop

cpu

uv sync --extra cpu

cuda

uv sync --extra cuda

dev

uv sync --extra dev --extra cpu

Models

Face Dtection

scrfd

Models accuracy on WiderFace benchmark:

Model	Easy	Medium	Hard
scrfd_10g_gnkps	95.51	94.12	82.14
scrfd_2.5g_gnkps	93.57	91.70	76.08
scrfd_500m_gnkps	88.70	86.11	63.57

来源https://github.com/SthPhoenix/InsightFace-REST/

runtime

• onnxruntime(cpu, cuda)
• ncnn
• opencv
• TensorRT

Face Landmark

datasets: Lapa134(Lapa106 + 28)

PFLD

项目来源：https://gold-one.coding.net/p/BeautyFace/d/TFace_Landmark/git 模型输出大小: $134\times 2$ runtime

• onnxruntime(cpu)
• ncnn
• TensorRT

rtmface

项目来源 https://blakeliu.coding.net/p/face/d/mmpose/git/tree/master/configs/face_2d_keypoint/rtmpose/lapa 模型输出大小: $134\times 2$

Model	NME
rtmpose-m-ort-lapa134	0.0288
rtmpose-s-ort-lapa134	0.0258

• onnxruntime(cpu, cuda)
• opencv
• ncnn
• TensorRT

Face Segmentation

ppliteseg

项目来源：https://github.com/tfrbt/FaceSeg.git

• label map 类别数量32

id	class
0	'background'
1	'skin'
2	'cheek'
3	'chin'
4	'ear'
5	'helix'
6	'lobule'
7	'bottom_lid'
8	'pupil'
9	'iris'
10	'sclera'
11	'tear_duct'
12	'top_lid'
13	'eyebrow'
14	'forhead'
15	'frown'
16	'hair'
17	'temple'
18	'jaw'
19	'beard'
20	'inferior_lip'
21	'oral comisure'
22	'superior_lip'
23	'teeth'
24	'neck'
25	'nose'
26	'ala_nose'
27	'bridge'
28	'nose_tip'
29	'nostril'
30	'DU26'
31	'sideburns'

Model	val IOU	val Dice
face_seg_ppliteseg_t	0.727	0.834

• onnxruntime(cpu)
• ncnn
• TensorRT

• label map 类别数量12

id	class
0	"background" # 0.背景
1	"skin" # 1.皮肤
2	"eye" # 2.眼睛
3	"pupil" # 3.瞳孔
4	"bottom_lid", # 4.下眼皮
5	"top_lid", # 5.上眼皮
6	"eyebrow" # 6.眉毛
7	"hair" # 7.头发
8	"superior_lip" # 8上嘴唇
9	"teeth" # 9.牙齿
10	"inferior_lip" # 10.下嘴唇
11	"nose" # 11.鼻子

Model	val IOU	val Dice
face12_seg_pplitesegb	0.774	0.869

• onnxruntime(cpu)
• ncnn
• TensorRT