Face_Detection_Challenging_Scenes in PyTorch

A PyTorch implementation of MFRNet: Face_Detection_Challenging_Scenes. Model size only 50M, We achieve better result with low computation FLOPS.

WiderFace Val Performance in single scale When using MFRNet as backbone net.

Style	easy	medium	hard
Pytorch (1200 to upto 1600)	94.85 %	93.50%	89.30%
Pytorch (original image scale)	95.48%	94.04%	85.43%

FDDB Performance.

FDDB(pytorch)	performance
MRFNet	98.64%

Contents

• Installation
• Training
• Evaluation
• TensorRT
• References

Installation

Clone and install

1. git clone https://github.com/yogesh0757/Face_Detection_Challenging_Scenes.git

2. Pytorch version 1.1.0+ and torchvision 0.3.0+ are needed.

3. Codes are based on Python 3

Data

1. Download the WIDERFACE dataset.

2. Download annotations (face bounding boxes & five facial landmarks) from baidu cloud or dropbox

3. Organise the dataset directory as follows:

  ./data/widerface/
    val/
      images/
      wider_val.txt

ps: wider_val.txt only include val file names but not label information.

Data1

We also provide the organized dataset we used as in the above directory structure.

Link: from google cloud or baidu cloud Password: ruck

Evaluation

Evaluation widerface val

1. Generate txt file

python test_widerface.py --trained_model weight_file --network mobile0.25 or resnet50

I used !python test_widerface.py --network resnet50 --cpu to generate the txt file. Then I downloaded the result in file.zip which contains the test result of validaion dataset in txt format. After this step I downloaded file.zip and moved it to widerface_evaluate/widerface_txt. 2. Evaluate txt results. Demo come from Here

cd ./widerface_evaluate
python setup.py build_ext --inplace
python evaluation.py

In order to plot the precision vs. recall curve I needed to make some modification in evaluation.py.Appended from line 274-277 as: plt.plot(recall, propose, color ='tab:blue') plt.savefig("save_curve.png",bbox_inches="tight") plt.xlabel("Recall") plt.ylabel("Precision") Before that I imported the python library matplotlib as plt. The precision vs. recall curve was saved in save_curve.png .

3. You can also use widerface official Matlab evaluate demo in Here

TensorRT

-TensorRT

References

• FaceBoxes
• Retinaface (mxnet)

@inproceedings{deng2019retinaface,
title={RetinaFace: Single-stage Dense Face Localisation in the Wild},
author={Deng, Jiankang and Guo, Jia and Yuxiang, Zhou and Jinke Yu and Irene Kotsia and Zafeiriou, Stefanos},
booktitle={arxiv},
year={2019}

REFERENCES: 1.https://drive.google.com/file/d/1iUYvk33zxV2dU-sG6EWUd4KW9YPzhc2n/view?usp=sharing -This is the google drive link of WIDER_VAL.zip which contains the validation images and can be accessed directly from my drive using this link. 2.https://drive.google.com/file/d/1rX11lpo3xyN8JsJ56r10KPu9IFALRQRX/view?usp=sharing -This is the google drive link of label.txt file of the validation dataset.It will be downloaded and saved as wider_val.txt file in the google colab notebook. 3.https://drive.google.com/file/d/1pLcsCaDSsfTrG01pm8WR5dCu0dJvzbKR/view?usp=sharing -This is the google drive link of file.zip which contains the result in txt format after testing the model with validation dataset. 4.https://colab.research.google.com/drive/1lReOo_0j50waqspBcE8zodT4jr3i9CJF?usp=sharing -This is the google drive link of my Retinaface_Pytorch.ipynb google colab notebook.My notebook can be directly accessed using this link.