Can EmotionNet be used with Transfer Learning to detect other facial emotions other than the orignal ones?

[raphaellcmb]

I would like to use EmotionNet as pretrained weights for fatigue detection with my own dataset, but in EmotionNet | NVIDIA NGC is written: “NVIDIA EmotionNet model were trained with MultiPie datasets that have six emotion classes. The subjects in the datasets were performing pose emotion. Therefore it will not be able to detect emotion classes that are not belong to these classes or spontaneous emotions.” Does this means that I can’t use EmotionNet to achieve my goal?

[Morganh]

You can use it as pretrained model to run training.

[raphaellcmb]

Thanks for the answer! Could you please provide me some examples of compatible models? In my understading the backbone model (pre trained) needs to be supported by the “target” network for the transfer learning. For example, you can use Resnet as a backbone for Detectnet_v2. So my question is, for which models can you use EmotionNet as backbone? Or I’d be using EmotionNet alone to do the training on the new dataset?

[Morganh]

@raphaellcmb

Or I’d be using EmotionNet alone to do the training on the new dataset?

Yes, you can run EmotionNet alone and also use EmotionNet | NVIDIA NGC to train on the new dataset. The EmotionNet constructs a network which is similar to resnet family.