wzk1015 / cnmt Goto Github PK

self.ocr_to_vocab = torch.ones((num_samples,50), device=torch.device("cuda"), dtype=torch.int64) * self.answer_processor.UNK_IDX
fwd_results["scores"][..., self.answer_processor.UNK_IDX] = -1e10

I cannot find it in your config file and search nothing on the Internet

作者您好，在按照requirments.txt文件安装时，总是提示requests==2.21.0版本与urllib3不匹配问题。请问这两者的版本分别是多少？python使用的是哪个版本？

i use your image dataset to train my model, but i cannot find ocr confidence in sample_list

(imdb_train.npy)

如题，object Faster R-CNN Features of TextCaps 特征挂掉了，网页链接没法下载

Thanks for your hard work!

I have such an error when I run the code that No module named 'pythia.models.cnmt'.

Have you ever met this error before? And could you help me to solve the error?

Thanks a lot!

作者您好，在按照requirments.txt文件安装时，总是提示requests==2.21.0版本与urllib3 botocore版本不匹配问题。请问这三者的版本分别是多少？python使用的是哪个版本？

Hello, thank you so much for sharing the code! Can the code of feature extraction be shared? Did you perform feature extraction on the original data set Textcaps?

I use your bash eval_val.sh and get the validation prediction, I submit it to the challenge but get the "CIDEr": 82.9。
I use your bash eval_test.sh and get the test prediction, I submit it to the challenge but get the "CIDEr": 76.54。
I wonder how can I get the right prediction?