thanks for your help

[vaeXu]

i download the dataset and try to train as the tutorial， i want to know how long does the program coast. and waht the following code will get.
Start training data server in the first terminal session. ./rmpe_dataset_server ../../keras_Realtime_Multi-Person_Pose_Estimation/dataset/train_dataset.h5 5555
Start validation data server in a second terminal session. ./rmpe_dataset_server ../../keras_Realtime_Multi-Person_Pose_Estimation/dataset/val_dataset.h5 5556

[murrayLuke]

It completely depends on your system, how powerful your gpu is, whether or not you have multiple gpu's. Also parameters such as the batch size can affect how long it takes as well. Without more information it's almost impossible to answer this. With a batch size of 5 and a gtx 970 it took between 48 and 72 hours to train. I started it Friday afternoon and it was done by Monday morning.

[vaeXu]

Thank you for your reply. I understand that rmpe_dataset_server is encapsulated. I can't see the code inside. How do you modify the code? Thank you very much again @murrayLuke

[anatolix]

It is here https://github.com/anatolix/keras_Realtime_Multi-Person_Pose_Estimation/tree/new-generation/py_rmpe_server