the speed on predicting a image or video about face_landmark_dnn HOT 5 CLOSED

Thanks @JWSunny
I didn't ported the model to android yet.
when I checked the model on macbook, the speed is average 39fps.

from face_landmark_dnn.

@junhwanjang thank you for your replay,
the speed 39fps is trained on basic models or mobilenets?

from face_landmark_dnn.

@JWSunny
The speed is
Mobilenets: 39fps
basic_models: about 42fps (but not accurate than mobilenets)

from face_landmark_dnn.

@junhwanjang OK, thanks.
how can i use the kersa_to_tf.py convert the Mobilenet_v1.hdf5 to pb file, can you provide an example including specific parameters; i find i can't use the hdf5 file in my conputer.
thank you very much, best wishes.

from face_landmark_dnn.

@junhwanjang
hello:
when i convert hdf5 to pb, the load_model(hdf5) is error: Process finished with exit code -1073740940 (0xC0000374), the main error is the model_config = f.attrs.get('model_config') in function load_model(); do you know why??

from face_landmark_dnn.