S³FD: Single Shot Scale-invariant Face Detector

A PyTorch Implementation of Single Shot Scale-invariant Face Detector

Description

Meanwhile train hand and head with S3FD,hand dataset is Egohands Dataset,head dataset is SCUT-HEAD,we can download hand model and face model

Requirement

• pytorch 0.3
• opencv
• numpy
• easydict

Prepare data

1. download WIDER face dataset、Egohands dataset and SCUT-HEAD
2. modify data/config.py
3. python prepare_wider_data.py 4 python prepare_handataset.py

Train

We can choose different dataset to train different target[face,head,hand]

python train.py --batch_size 4 --dataset face\hand\head

Evalution

according to yourself dataset path,modify data/config.py

1. Evaluate on AFW.

python afw_test.py

2. Evaluate on FDDB

python fddb_test.py

3. Evaluate on PASCAL face

python pascal_test.py

4. test on WIDER FACE

python wider_test.py

Demo

you can test yourself image

python demo.py

Result

1. AFW PASCAL FDDB

AFW AP=99.81 paper=99.85 
PASCAL AP=98.77 paper=98.49
FDDB AP=0.975 paper=0.983
WIDER FACE:
Easy AP=0.925 paper = 0.927
Medium AP=0.925 paper = 0.924
Hard AP=0.854 paper = 0.852

2. demo

References

• S³FD: Single Shot Scale-invariant Face Detector
• ssd.pytorch