Skip to content

Latest commit

 

History

History
69 lines (52 loc) · 4.17 KB

README.md

File metadata and controls

69 lines (52 loc) · 4.17 KB

English | 简体中文

SCRFD Ready-to-deploy Model

Export ONNX Model

# Download scrfd model files
e.g. download from  https://onedrive.live.com/?authkey=%21ABbFJx2JMhNjhNA&id=4A83B6B633B029CC%215542&cid=4A83B6B633B029CC

# Install the official library to configure the environment. This version should be exported in the following environment:
- Configure the environment manually
  torch==1.8.0
  mmcv==1.3.5
  mmdet==2.7.0

- Configure via docker
  docker pull qyjdefdocker/onnx-scrfd-converter:v0.3

# Export files in onnx format
- Manual generation
  python tools/scrfd2onnx.py configs/scrfd/scrfd_500m.py weights/scrfd_500m.pth --shape 640 --input-img face-xxx.jpg

- docker
  onnx files are in docker's onnx directory

Download Pre-trained ONNX Models

For developers' testing, models exported by SCRFD are provided below. Developers can download and use them directly. (The accuracy of the models in the table is sourced from the official library)

Model Size Accuracy
SCRFD-500M-kps-160 2.5MB -
SCRFD-500M-160 2.2MB -
SCRFD-500M-kps-320 2.5MB -
SCRFD-500M-320 2.2MB -
SCRFD-500M-kps-640 2.5MB 90.97%
SCRFD-500M-640 2.2MB 90.57%
SCRFD-1G-160 2.5MB -
SCRFD-1G-320 2.5MB -
SCRFD-1G-640 2.5MB 92.38%
SCRFD-2.5G-kps-160 3.2MB -
SCRFD-2.5G-160 2.6MB -
SCRFD-2.5G-kps-320 3.2MB -
SCRFD-2.5G-320 2.6MB -
SCRFD-2.5G-kps-640 3.2MB 93.8%
SCRFD-2.5G-640 2.6MB 93.78%
SCRFD-10G-kps-320 17MB -
SCRFD-10G-320 15MB -
SCRFD-10G-kps-640 17MB 95.4%
SCRFD-10G-640 15MB 95.16%
SCRFD-10G-kps-1280 17MB -
SCRFD-10G-1280 15MB -

Detailed Deployment Tutorials

Release Note