PyramidBox

This is an unofficial Tensorflow re-implementation of PyramidBox: A Context-assisted Single Shot Face Detector, which achieves superior performance among the state-of-the-art on the two common face detection benchmarks, FDDB and WIDER FACE.

Note

There is still a gap in performance from the paper. May be caused by several reasons:

  • Without implementing data-anchor-sampling.
  • Differences of data augmentation from original.
  • The batch size in the paper is 16, but I used 1 because of the limitation of memory.
  • Hyperparameters not mentioned in the paper.
  • Differences of deep learning framework.

Results

Face Detection


Results on WIDER FACE validation set:

This is just a very casual training result. I believe you can achieve better results after trying some other hyperparameters. For example: batch size, learning rate and some parameters related to the loss function,etc.

Method AP Easy AP Medium AP Hard
original 96.1 95.0 88.9
this repo 90.6 88.8 73.4

Usage

Prerequisites

(Only tested on) Ubuntu 16.04 with:

  • Python3
  • Tensorflow-gpu 1.4
  • OpenCV3

Clone the repo

git clone https://github.com/EricZgw/PyramidBox.git
python makedir.py

Download PyramidBox models form BaiduYun or GoogleDrive .

Demo

Run the following script for visualization:

python demo.py

Train on WIDER FACE Datasets

  1. Download pre-trained VGG16 models from here and put it to /checkpoints.
  2. Download WIDER FACE Datasets and convert to VOC format. Path looks like below:
datasets/
       |->widerface/
       |    |->WIDER_train/
       |    |->WIDER_val/
       |    |->WIDER_test/
       |    |->Annotations/
       |    |->JPEGImages/
       |    |...
  1. Run the following script to generate TFRecords:
python datasets/pascalvoc_to_tfrecords.py
You can run `check_data_io.py` to check data. This step is not necessary.
  1. The training strategy is two-stages:
    First run train_model.py with below setting to train additional PyramidBox layers:
self.fine_tune_vgg16 = False
  1. Then set self.fine_tune_vgg16 =Ture to run train_model.py to train total network.

Validation

Run the following script for evaluation and get mAP:

python widerface_eval.py
cd eval/eval_tools
octave wider_eval.m

TODO

  • Add data-anchor-sampling
  • Try more logical and rigorous data augmentation
  • Transfer to other backbone networks

GitHub