Sha-Lab / CMHSE

The code repository for "Cross-Modal and Hierarchical Modeling of Video and Text" in PyTorch
16 stars 6 forks source link

Extracting the features of a new dataset #2

Open carlamao opened 3 years ago

carlamao commented 3 years ago

Hi, I am trying to follow COOT's implementation using a different dataset, ActivtiyNet-Entities. They used your model to extract the features. What steps should I take to do so?

DesaleF commented 3 years ago

@carlamao I am trying to do the same for COOT video captioning, But I don't know the steps. Did you succeed in extracting the features. Can you share me the steps you followed if you get any success?

carlamao commented 3 years ago

I used the feature provided with ActivityNet Entities and manually modified the annotation files because I couldn't find a way to compute those features.