search

FaceONNX / FaceONNX

Face recognition and analytics library based on deep neural networks and ONNX runtime
MIT License
189 stars 33 forks source link

Do you have benchmark about the model? #20

Closed stereomatchingkiss closed 10 months ago

stereomatchingkiss commented 10 months ago

Like LFW, megaface, mscelab? Thanks

asiryan commented 10 months ago

@stereomatchingkiss hello! What model you are talking about?

stereomatchingkiss commented 10 months ago

@stereomatchingkiss hello! What model you are talking about?

Mainly recognition_resnet27.onnx, more precisely, benchmark of the face recognition pipeline of this project. Thanks

asiryan commented 10 months ago

@stereomatchingkiss hello! What model you are talking about?

Mainly recognition_resnet27.onnx, more precisely, benchmark of the face recognition pipeline of this project. Thanks

You can find it here https://github.com/huangyangyu/SeqFace

asiryan commented 10 months ago

In my local tests this is a more accurate and strong model

stereomatchingkiss commented 10 months ago

In my local tests this is a more accurate and strong model

Nice to hear that and thanks for the benchmark, I think this project will become more attractive if with benchmark. Any plan to release light weight model?

asiryan commented 10 months ago

In my local tests this is a more accurate and strong model

Nice to hear that and thanks for the benchmark, I think this project will become more attractive if with benchmark. Any plan to release light weight model?

I think no. But I want to release project for Person Recognition and Re-identification

stereomatchingkiss commented 10 months ago

In my local tests this is a more accurate and strong model

Nice to hear that and thanks for the benchmark, I think this project will become more attractive if with benchmark. Any plan to release light weight model?

I think no. But I want to release project for Person Recognition and Re-identification

Nice to hear that, I am writing an open source face recognition webapp by Qt6 and face-recognition-ncnn.

The results of insight face looks great and it has small model which suit for webapp(this webapp run on local machine, not suit for big model), but the pretrained model of insight face cannot use for commercial purpose. I am finding alternative solution

asiryan commented 10 months ago

In my local tests this is a more accurate and strong model

Nice to hear that and thanks for the benchmark, I think this project will become more attractive if with benchmark. Any plan to release light weight model?

I think no. But I want to release project for Person Recognition and Re-identification

Nice to hear that, I am writing an open source face recognition webapp by Qt6 and face-recognition-ncnn.

The results of insight face looks great and it has small model which suit for webapp(this webapp run on local machine, not suit for big model), but the pretrained model of insight face cannot use for commercial purpose. I am finding alternative solution

Wow, send me a link to these models, I'll check it out.

stereomatchingkiss commented 10 months ago

In my local tests this is a more accurate and strong model

Nice to hear that and thanks for the benchmark, I think this project will become more attractive if with benchmark. Any plan to release light weight model?

I think no. But I want to release project for Person Recognition and Re-identification

Nice to hear that, I am writing an open source face recognition webapp by Qt6 and face-recognition-ncnn. The results of insight face looks great and it has small model which suit for webapp(this webapp run on local machine, not suit for big model), but the pretrained model of insight face cannot use for commercial purpose. I am finding alternative solution

Wow, send me a link to these models, I'll check it out.

Here. More benchmark and models link, but they do not release the mobileNet model

asiryan commented 9 months ago

In my local tests this is a more accurate and strong model

Nice to hear that and thanks for the benchmark, I think this project will become more attractive if with benchmark. Any plan to release light weight model?

I think no. But I want to release project for Person Recognition and Re-identification

Nice to hear that, I am writing an open source face recognition webapp by Qt6 and face-recognition-ncnn. The results of insight face looks great and it has small model which suit for webapp(this webapp run on local machine, not suit for big model), but the pretrained model of insight face cannot use for commercial purpose. I am finding alternative solution

Wow, send me a link to these models, I'll check it out.

Here. More benchmark and models link, but they do not release the mobileNet model

Thank you!