MuKEA
MuKEA copied to clipboard
MuKEA: Multimodal Knowledge Extraction and Accumulation for Knowledge-based Visual Question Answering
Hi, why have you only used the train and test split among image data for KR-VQA? you always skipped the validation data, and it is not even in your krvqa.zip...
Hi, Could you please upload your model for Krisp+mukea as it is your best result? thx.
您好,可以提供krvqa数据集的validation set的特征krvqa_img_feature_val.pickle吗,已下载的特征里只有train和test的
您好,datasets.py文件中,args.dataset参数不同会使用不同的预训练数据集,请问vqa_train_filter.json和vqa_train.json有什么不同,当args,dataset==vqav2时,会将vqa_img_feature_train.pickle和vqa_img_feature_val.pickle合并起来做训练,请问您在论文中报告的实验,预训练时具体是用哪种组合呢?比如:pretrain时datasets是vqav2,不做validate, finetune时用okvqa或krvqa
i want to download this ,But there are always Failures - Forbidden Download annotation from [google drive](https://drive.google.com/file/d/1YuOUTbK7rged0gopQko5rdQuHKxiW9sv/view?usp=sharing) how can I to get it
大佬,我在用mukea训练自己的数据集时报错,没有MuKEA/data/kr-vqa/krvqa-ans_dic.pickle这个文件,我看您readme中也没有写,这个文件是如何色生成的,有什么用啊?