face detection SSD: Single Shot MultiBox Object Detector for face detection

SSD is an unified framework for object detection with a single network. caffe base on https://github.com/weiliu89/caffe.git.Making some small changes for supporting python3.x

Train the model

This example only covers training on Pascal VOC format dataset.

• Download the converted pretrained VGG_ILSVRC_16_layers_fc_reduced.caffemodel model.
• Compile the caffe_ssd,see https://github.com/weiliu89/caffe.git for details.
• Download the your Pascal VOC format dataset into data/
• modify tools/create_list.shandtools/create_data.sh to your own parameters, and run to get the lmdbdataset.
• modify train/ssd_config.pyto your own parameters, run to get the .rec files for train.
• run train/ssd_train.shfor training.

Demo results

Try the demo

Both python and c++ demo are provided.

#for c++
cd demo/cpp
mkdir build
cd build && cmake .. && make && cd ..
./build/demo ../../images

#for python
python3 demo/python/ssd_detect_face.py