ly19965 / damofd

Introduction

DamoFD introduces a novel DDSAR-score to characterize backbone stage-level detection ability, based on which and SCRFD detection framework, a sota family-based backboen architectures are searched automatically. Now, our paper is accepted by ICLR-2023.

Quickly Start on EasyFace。

Performance

Precision, flops are both evaluated on VGA resolution.

Installation

Please refer to mmdetection for installation.

1. Install mmcv. (mmcv-full==1.2.6 and 1.3.3 was tested)
2. Install build requirements and then install mmdet.

   pip install -r requirements/build.txt
   pip install -v -e .  # or "python setup.py develop"
   

Data preparation

WIDERFace:

1. Download WIDERFace datasets and put it under data/retinaface.
2. Download annotation files from gdrive and put them under data/retinaface/

  data/retinaface/
      train/
          images/
          labelv2.txt
      val/
          images/
          labelv2.txt
          gt/
              *.mat
          

Annotation Format

please refer to labelv2.txt for detail

For each image:

# <image_path> image_width image_height
bbox_x1 bbox_y1 bbox_x2 bbox_y2 (<keypoint,3>*N)
...
...
# <image_path> image_width image_height
bbox_x1 bbox_y1 bbox_x2 bbox_y2 (<keypoint,3>*N)
...
...

Keypoints can be ignored if there is bbox annotation only.

Training

Example training command, with 4 GPUs:

CUDA_VISIBLE_DEVICES="0,1,2,3" PORT=29701 bash ./tools/dist_train.sh ./configs/DamoFD/DamoFD_0.5g.py 4

WIDERFace Evaluation

We use a pure python evaluation script without Matlab.

GPU=0
GROUP=scrfd
TASK=scrfd_2.5g
CUDA_VISIBLE_DEVICES="$GPU" python -u tools/test_widerface.py ./configs/"$GROUP"/"$TASK".py ./work_dirs/"$TASK"/model.pth --mode 0 --out wouts

Pretrained-Models

Name	Easy	Medium	Hard	FLOPs	Params(M)	Link
DamoFD_0.5G	90.32	88.36	71.03	520M	0.26	download
DamoFD_2.5G	92.82	91.48	78.70	2.46G	0.44	download
DamoFD_10G	95.14	94.29	84.07	9.74G	1.27	download
DamoFD_34G	95.63	94.80	85.08	34.03G	6.24	download
DamoFD_0.5G_KPS	90.41	88.45	69.97	593M	0.27	EasyFace
DamoFD_2.5G_KPS	93.25	91.85	78.17	2.53G	0.45	EasyFace
DamoFD_10G_KPS	95.42	94.35	83.94	9.83G	1.28	EasyFace
DamoFD_34G_KPS	95.96	95.08	86.09	34.32G	6.28	EasyFace

mAP, FLOPs are all evaluated on VGA resolution. _KPS means the model includes 5 keypoints prediction.

Convert to ONNX

Please refer to tools/scrfd2onnx.py

Generated onnx model can accept dynamic input as default.

You can also set specific input shape by pass --shape 640 640, then output onnx model can be optimized by onnx-simplifier.

Inference

Please refer to tools/scrfd.py which uses onnxruntime to do inference.

Network Search

cd DDSAR
sh scripts/DamoFd_500M.sh

As mentioned in paper, we set α to 0.25 when searching DDSAR-500M models. While for DDSAR-2.5G, DDSAR-10G and DDSAR-34G models, we need to find a suitable value of α and add some constraints into search space. Otherwise, some trivial architectures may be searched.

Acknowledgments

We thank Insighface for the excellent code base.