jiangxiluning / facenet_mtcnn_to_mobile

convert facenet and mtcnn models from tensorflow to tensorflow lite and coreml (使用 TFLite 将 FaceNet 和 MTCNN 移植到移动端)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

请问,MTCNN的三个.tflite如何使用?

Annie201521010625 opened this issue · comments

请问,MTCNN的三个.tflite如何使用?

@jiangxiluning : Can you plz specify how to use the three .tflite files and what is the difference between them?

Hello, as far as I know, the MTCNN detection pipeline involves 3 stages (pnet, rnet, onet). pnet does some rough processing and hands over the results (areas in the input image where faces are potentially present) to rnet. rnet further refines the results and pass them off to onet. Onet finally determines which of the proposed regions contains actual faces, and outputs the bounding box and face probabilities.
So, you have to actually implement the MTCNN 3-stage pipeline and use the 3 .tflite files to load parameters into pnet, rnet, and onet (which isn't really a trivial process)