search

ruc-aimc-lab / LAFF

Source code of ECCV2022 LAFF for Text-to-Video Retrieval
43 stars 4 forks source link

Data download #5

Open aranciokov opened 10 months ago

aranciokov commented 10 months ago

Hi, thanks for sharing this incredible work! I'm trying to download the data, but it seems nearly impossible. Is it possible to upload to somewhere different from pan.baidu? e.g., google drive, dropbox, or something like that

Thanks! Best regards, Alex

fly-dragon211 commented 10 months ago

Hi, thanks for sharing this incredible work! I'm trying to download the data, but it seems nearly impossible. Is it possible to upload to somewhere different from pan.baidu? e.g., google drive, dropbox, or something like that

Thanks! Best regards, Alex

It is the googlde drive: https://drive.google.com/drive/folders/1A2f3qtBsNd1cbOkwOke0NpjAuItyM6K9?usp=sharing

aranciokov commented 10 months ago

Hi, thank you for your reply. I'm trying to run some of the shell scripts, e.g. this one: do_laffml_mvtest3k.sh. However, it seems like it also requires the features for mean_C3d_resneXt101_16f, which I can not find for any of the datasets. Is it possible to obtain those as well?

Thanks! Best, Alex

aranciokov commented 10 months ago

Hi, I forgot to mention in the previous reply that also the text feature data are missing. For instance, the code looks for this file: VisualSearch/word2vec/flickr/vec500flickr30m/shape.txt but the word2vec folder does not exist.

fly-dragon211 commented 9 months ago

Hi, I forgot to mention in the previous reply that also the text feature data are missing. For instance, the code looks for this file: VisualSearch/word2vec/flickr/vec500flickr30m/shape.txt but the word2vec folder does not exist.

We have revised the files on Google Drive. Please try accessing them again. We look forward to your response! https://drive.google.com/drive/folders/1A2f3qtBsNd1cbOkwOke0NpjAuItyM6K9?usp=sharing

aranciokov commented 9 months ago

Hi, thank you for your help! I managed to reproduce similar results on most of the datasets, which is great. I'm currently having issues with VATEX, for which the features seem to have different shapes. In particular, looking at the shape.txt files:

but, HowTo100M_TimeSformer_divST_96x4_224 -> 1970 768. This seems to imply the uploaded file/folder of TimeSformer for VATEX is not the correct one.

Thanks again for your help!

fly-dragon211 commented 8 months ago

Hi, thank you for your help! I managed to reproduce similar results on most of the datasets, which is great. I'm currently having issues with VATEX, for which the features seem to have different shapes. In particular, looking at the shape.txt files:

  • clip_finetune_8frame_uniform_1103 -> 32239 512
  • mean_irCSN_152_ig65m_from_scratch -> 32239 2048
  • X3D_L -> 32239 2048

but, HowTo100M_TimeSformer_divST_96x4_224 -> 1970 768. This seems to imply the uploaded file/folder of TimeSformer for VATEX is not the correct one.

Thanks again for your help!

Thank you for your correction, we have updated the files. https://drive.google.com/drive/folders/1A2f3qtBsNd1cbOkwOke0NpjAuItyM6K9?usp=sharing