search

piergiaj / super-events-cvpr18

Code for our CVPR 2018 paper "Learning Latent Super-Events to Detect Multiple Activities in Videos"
MIT License
123 stars 34 forks source link

what is the mask #9

Open WestBbbbrook opened 4 years ago

WestBbbbrook commented 4 years ago

I didn't find the mask in the dataset，can you tell me what is it？

 # forward
    if not baseline:
        outputs = model([inputs, torch.sum(mask, 1)])
    else:
        outputs = model(inputs)
kkahatapitiya commented 4 years ago

Look at the collate function in dataset file. When batch_size>1, it pads temporally small input features in a batch to a common temporal size (max). Mask is 1 where for actual duration of the padded input feature, and 0 otherwise.