search

deepinsight / insightface

State-of-the-art 2D and 3D Face Analysis Project
https://insightface.ai
22.17k stars 5.28k forks source link

Use retinaface to face detection alignment #1018

Open tomriddle54 opened 4 years ago

tomriddle54 commented 4 years ago

Hello @yangfly @vizakshat @MyraBaba @yingfeng @nttstar I have change the code detection code from mtcnn to retina face ret = self.detector.detect_face(face_img, det_type = self.args.det) to retina face detector.detect(img, thresh, scales=scales, do_flip=flip) but when nimg = face_preprocess.preprocess(face_img, bbox, points1, image_size='112,112') is called i am not getting the proper face i am getting full frame insterd of crop align face is there anything I am missing.

brealisty commented 3 years ago

@tomriddle54 have resolved this problem？

brealisty commented 3 years ago

Hello @yangfly @vizakshat @MyraBaba @yingfeng @nttstar I have change the code detection code from mtcnn to retina face ret = self.detector.detect_face(face_img, det_type = self.args.det) to retina face detector.detect(img, thresh, scales=scales, do_flip=flip) but when nimg = face_preprocess.preprocess(face_img, bbox, points1, image_size='112,112') is called i am not getting the proper face i am getting full frame insterd of crop align face is there anything I am missing.

don't know whether resolved this problem, I get the right output, this function's parameters should be like this: (img, face_bbox, landmark, image_size='112,112') img is full image not face_img, face_box is 1-D(cordinate for 1 face), landmark is 2-D(cordinate for 1 face)