insightface/recognition/arcface_paddle/README.md

简体中文 | English

Arcface-Paddle

1. Introduction

Arcface-Paddle is an open source deep face detection and recognition toolkit, powered by PaddlePaddle. Arcface-Paddle provides three related pretrained models now, include BlazeFace for face detection, ArcFace and MobileFace for face recognition.

• This tutorial is mainly about face recognition.
• For face detection task, please refer to: Face detection tuturial.
• For Whl package inference using PaddleInference, please refer to whl package inference.

Note: Many thanks to GuoQuanhao for the reproduction of the Arcface basline using PaddlePaddle.

2. Environment preparation

Please refer to Installation to setup environment at first.

3. Data preparation

3.1 Download dataset

Download the dataset from insightface datasets.

• MS1M_v2: MS1M-ArcFace
• MS1M_v3: MS1M-RetinaFace

3.2 Extract MXNet Dataset to images

python tools/mx_recordio_2_images.py --root_dir ms1m-retinaface-t1/ --output_dir MS1M_v3/

After finishing unzipping the dataset, the folder structure is as follows.

MS1M_v3
|_ images
|  |_ 00000001.jpg
|  |_ ...
|  |_ 05179510.jpg
|_ label.txt
|_ agedb_30.bin
|_ cfp_ff.bin
|_ cfp_fp.bin
|_ lfw.bin

Label file format is as follows.

# delimiter: "\t"
# the following the content of label.txt
images/00000001.jpg 0
...

If you want to use customed dataset, you can arrange your data according to the above format.

4. How to Training

4.1 Single node, single GPU

Dynamic Mode

python -m paddle.distributed.launch --gpus=0 tools/train.py \
    --config_file configs/ms1mv2_mobileface.py \
    --is_static False \
    --embedding_size 128 \
    --sample_ratio 1.0 \
    --loss ArcFace \
    --batch_size 512 \
    --dataset MS1M_v2 \
    --num_classes 85742 \
    --data_dir MS1M_v2/ \
    --label_file MS1M_v2/label.txt \
    --fp16 False

4.2 Single node, 8 GPUs:

Static Mode

sh scripts/train_static.sh

Dynamic Mode

sh scripts/train_dynamic.sh

During training, you can view loss changes in real time through VisualDL, For more information, please refer to VisualDL.

5. Model evaluation

The model evaluation process can be started as follows.

Static Mode

sh scripts/validation_static.sh

Dynamic Mode

sh scripts/validation_dynamic.sh

6. Export model

PaddlePaddle supports inference using prediction engines. Firstly, you should export inference model.

Static Mode

sh scripts/export_static.sh

Dynamic Mode

sh scripts/export_dynamic.sh

We also support export to onnx model, you only need to set --export_type onnx.

7. Model inference

The model inference process supports paddle save inference model and onnx model.

sh scripts/inference.sh

8. Model performance

8.1 Performance of Lighting Model

Configuration：

• CPU: Intel(R) Xeon(R) Gold 6184 CPU @ 2.40GHz
• GPU: a single NVIDIA Tesla V100
• Precison: FP32
• BatchSize: 64/512
• SampleRatio: 1.0
• Embedding Size: 128
• MS1MV2

Model structure	lfw	cfp_fp	agedb30	CPU time cost	GPU time cost	Inference model
MobileFace-Paddle	0.9952	0.9280	0.9612	4.3ms	2.3ms	download link
MobileFace-mxnet	0.9950	0.8894	0.9591	7.3ms	4.7ms	-

• Note: MobileFaceNet-Paddle training using MobileFaceNet_128

8.2 Accuracy on Verification Datasets

Configuration：

• GPU: 8 NVIDIA Tesla V100 32G
• Precison: Pure FP16
• BatchSize: 128/1024

Mode	Datasets	backbone	Ratio	agedb30	cfp_fp	lfw	log	checkpoint
Static	MS1MV3	r50	0.1	0.98317	0.98943	0.99850	log	checkpoint
Static	MS1MV3	r50	1.0	0.98283	0.98843	0.99850	log	checkpoint
Dynamic	MS1MV3	r50	0.1	0.98333	0.98900	0.99833	log	checkpoint
Dynamic	MS1MV3	r50	1.0	0.98317	0.98900	0.99833	log	checkpoint

8.3 Maximum Number of Identities

Configuration：

• GPU: 8 NVIDIA Tesla V100 32G (32510MiB)
• BatchSize: 64/512
• SampleRatio: 0.1

Mode	Precision	Res50	Res100
Framework1 (static)	AMP	42000000 (31792MiB)	39000000 (31938MiB)
Framework2 (dynamic)	AMP	30000000 (31702MiB)	29000000 (32286MiB)
Paddle (static)	Pure FP16	60000000 (32018MiB)	60000000 (32018MiB)
Paddle (dynamic)	Pure FP16	59000000 (31970MiB)	59000000 (31970MiB)

Note: config environment variable export FLAGS_allocator_strategy=naive_best_fit

8.4 Throughtput

Configuration：

• BatchSize: 128/1024
• SampleRatio: 0.1
• Datasets: MS1MV3
• V100: Driver Version: 450.80.02, CUDA Version: 11.0
• A100: Driver Version: 460.32.03, CUDA Version: 11.2

For more experimental results see PLSC, which is an open source Paddle Large Scale Classification Tools powered by PaddlePaddle. It supports 60 million classes on 8 NVIDIA V100 (32G).

9. Inference using PaddleInference

9.1 Install insightface-paddle

# install insightface-paddle
pip install --upgrade insightface-paddle

9.2 Download the index gallery, demo image.

mkdir -p images/gallery/
mkdir -p images/query/

# Index library for the recognition process
wget https://raw.githubusercontent.com/littletomatodonkey/insight-face-paddle/main/demo/friends/index.bin -P images/gallery/
# Demo image
wget https://raw.githubusercontent.com/littletomatodonkey/insight-face-paddle/main/demo/friends/query/friends2.jpg -P images/query/

9.3 Inference using MobileFace

# default using MobileFace
insightfacepaddle \
    --det \
    --rec \
    --index=images/gallery/index.bin \
    --input=images/query/friends2.jpg \
    --output="./output"

The final result is save in folder output/, which is shown as follows.

For more details about parameter explanations, index gallery construction and whl package inference, please refer to Whl package inference tutorial.