search

LAHAproject / LatentGoal

Code accompanying the IEEE WACV 2022 paper "Action anticipation using latent goal learning"
MIT License
1 stars 1 forks source link

About the implementation of the 50Salads and Breakfast datasets #1

Closed Jennifer123www closed 2 years ago

Jennifer123www commented 2 years ago

Thank you for sharing. I would like to ask, where is the code for the specific implementation of 50 Salads and Breakfast mentioned in your paper? What special treatments do these two datasets require?

debadityaroy commented 2 years ago

Hi,

Thanks for your interest. The I3D features were obtained from this repo for both 50Salads and Breakfast.

https://github.com/yabufarha/ms-tcn

Download the data folder, which contains the features and the ground truth labels. (~30GB) (If you cannot download the data from the previous link, try to download it from here)

Jennifer123www commented 2 years ago

As I read your paper, I noticed that the code you gave was only the processing and implementation of the EPIC-KITCHENS55 dataset. I noticed that the data organization format for the 50Salads and Breakfast datasets was not the same as the format of the EPIC-KITCHENS55 dataset, for example, the 50Salads and Breakfast datasets did not have separate comments on verbs, nouns, etc., and I was confused by the treatment of these two datasets. Is it convenient for you to publish the implementation of this code? Also, what is the meaning of the word "obs_seg" in the code? I look forward to hearing from you.

debadityaroy commented 2 years ago

Hi

  1. 'obs_seg' refers to observed segment or observed duration.
  2. For the 50Salads and Breakfast dataset, you will just have actions instead of verbs or nouns.
  3. I have posted the code for training on Breakfast dataset using I3D features.https://github.com/debadityaroy/LatentGoal/blob/main/i3d_latent_goal_bf.py