Closed MrShouxingMa closed 11 months ago
Hi, @MrShouxingMa The visual feature is used without modification in our model, you may download baby image feature at: http://snap.stanford.edu/data/amazon/productGraph/image_features/categoryFiles/image_features_Baby.b
Thanks for your feedback.
@MrShouxingMa By the way, all raw datasets are located at: http://jmcauley.ucsd.edu/data/amazon/links.html
However, the paper may automatically redirect to Version 2018.
To access the 2014 version of the dataset from the paper, please follow these steps:
esc
key repeatedly until the page stops trying to redirect to the 2018 version.Please note that timing is crucial in this process to prevent the automatic redirection. If you're redirected to the 2018 version, simply go back and try again.
Thank you very much for your prompt and patient reply!
With the guidance of your detailed tutorial, I found out by downloading the raw data that indeed there are many zero vectors in the raw data, which is similar to the data in your paper.
Thank you very much for sharing the multi-modal dataset!
When I looked at the Baby dataset in detail, I found that its visual features were a bit strange compared to its textual features. According to the article description, I found the EMNLP-IJCNLP 2019 article and wanted to download and view its visual data. It was a pity that the data no longer supports downloading.
Therefore, I would like to confirm with you the accuracy of its data. The visual features of the Baby dataset are as follows,
While the text features of the Baby dataset are as follows,
Best wishes