Skip to content

ansaricard/MTCNN_YOLO11

Repository files navigation

MTCNN_YOLO11

combine onet from MTCNN with yolo11 face detection wider face

pip install ultralytics
pip install onnx onnxruntim

you can use wider-face pretrained model yolov11 as face detection then for face recognition, you need face alignment, here is the best location, so follow me

the challenge:

mtcnn have 3 part: pnet, rnet, onet: at last step ie onet, landmark is the output of network so: input: face cropted result from yolo11 (trained on wider face) model: accept 1,3,48,48 rgb format float32 so after preprocss give the input to onet, then post process the output for alignment wiht simple image processing trasformation on image.

how to use:

python main_onnx.py #for run onet in onnx format, recommened (do not need to pytorch!)
python main.py #for run onet in torch format (same as MTCNN_pytorch release)

you can change the input image path in main.py to run model on different images I am sure it is a submodule in a big project, so no need to install this project as a package , .. just use these modules

About

combine onet from MTCNN with yolo11 face detection wider face, just landmark then alingment

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages