Training with confidence scores

[anas-zafar]

Hi @BrianG13 @sigal-raab you mentioned in the paper that "OpenPose and Iskakov et al. provide confidence values that we add to the network input. CPN does not provide these values, hence we assign identical confidence values for all joints when using it"

Can you please confirm the weights that you have provided is trained using confidence score? If yes, could you please guide me where should I pass the confidence scores for testing in the wild. If no, can you please guide me how can I retrain the model using confidence scores?

Thanks lots

[BrianG13]

Hi @anas-zafar

Our published model weights is without using the confidence score, but if you want to train using those it is very simple.

Just change the config json here: https://github.com/BrianG13/FLEX/blob/main/config_zoo/default.json#L18 to True and it should work :)

What is the data you are using to train your model?

[anas-zafar]

I am trying to train it using open pose keypoints. Can you please share the file structure you use it as an input for training the model with the confidence scores? @BrianG13

[anas-zafar]

It seems there is no declaration of self.add_noise

if config.arch.confidence:
            self.poses_2d_noised, confidence_maps = self.add_noise(self.poses_2d, training=is_train)