Model used for feature extraction in Temporal Action Localization

[christian-matroid]

Hello! I have been able to successfully use the features posted in Temporal Action Localization to reproduce great results on their respective datasets. Are you planning to release the weights of the models that generated these features?

In addition, could you explain the training + finetuning process for these models? Was each dataset feature generated by a model that was finetuned on that dataset?