libeineu / MMT-VQA

This repository is the official implement of our EMNLP paper "Incorporating Probing Signals into Multimodal Machine Translation via Visual Question-Answering Pairs" (Findings)
MIT License
7 stars 0 forks source link

How to get the "data-bin/multi30k.en-de.mmt_vqa/train.imgname" #1

Closed Jianglllri closed 1 month ago

Jianglllri commented 5 months ago

Hi Thank you for your great work. I found No such file or directory when trying to run your code: 'data-bin/multi30k.en-de.mmt_vqa/train.imgname'.I'd like to know what code was executed to generate this file?

yuxinzuo commented 4 months ago

Apologies for omitting a description of the source of this file.

'data-bin/multi30k.en-de.mmt_vqa/train.imgname' indicates image file name of multi30k, you can get it from here: https://github.com/multi30k/dataset/blob/master/data/task1/image_splits/train.txt.

Note: You can post-process this file based on the paths of the Flickr images you downloaded.