search

microsoft / UniVL

An official implementation for " UniVL: A Unified Video and Language Pre-Training Model for Multimodal Understanding and Generation"
https://arxiv.org/abs/2002.06353
MIT License
339 stars 54 forks source link

Can you share your HowTo100M.csv file? #30

Open ShinJQ opened 2 years ago

ShinJQ commented 2 years ago

Hi! From your paper and readme.md file on (https://github.com/microsoft/UniVL)/dataloaders/, I could infer that the csv file you've used differ from the original csv file.

It is mentioned that 1.2M videos are used for pretraining. Can you share the csv file that contains the id of1.2M video you've used for pretraining?

ArrowLuo commented 2 years ago

Hi @ShinJQ, I am afraid that I can not share the file now. The official CSV contains about 1.2M video ids so you can generate the HowTo100M.csv easily. Best.

HuBot2020 commented 2 years ago

Hi @ArrowLuo , sorry to bother, but I'm noticing that the pre-extracted feature files for HowTo100M supplied by the dataset owner have a file extension of '.mp4.npy' while in the README for uniVL you have the feature file with file extension '.npy'. Is this of any concern? Did you and your team do any extra processing for the feature files to get the '.npy' file extension for feature files?

ArrowLuo commented 2 years ago

Hi @HuBot2020, It is ok for the postfix of the feature filename, e.g., '.mp4.npy' or '.npy'. We have no other processing on the extracted feature. But I do not know what the difference is between the '.mp4.npy shared by the dataset owner and ours.