soloist97/densecap-pytorch

A simplified pytorch version of densecap

densecap-pytorch's Issues

License missing

Hi there!

First of all thank you for sharing this implementation!
I would like to use this Implementation for my master thesis, but I noticed that there is no license information attached. Would you mind adding this to your repository?

How to get pretrained model?

Hello.

I'm glad I found your nice work! Thank you.
I am trying to download pre-trained model as outlined in Readme.

1. There are two pth.tar files when I access to the OneDrive. Which one is a pretrained model? And what is the difference between them?

2. When I downloaded them, there are no files in pth.tar files like below screenshot. Could you please check if that file is still available?
   

How to visualize the ground truth of dense captioning?

Hello, why do I encounter the following error when I want to use visual image to text vg_dataset_visualization.ipynb? Is there any solution? The VG data set format is also set according to the format you showed.

About pointed box to produce its caption

Dear scholar,
I am glad to read your code about densecaptioning. And I want to know if the code can realize that I pass the designated boxes to the net, and the net return the specific boxes' captions?

How to unzip pretrained model

Hi,
I download the trained model train_all_val_all_bz_2_epoch_10_inject_init.pth.tar from one drive. Also, I want to use this to evaluate on my own image with describe.py.

1. Can you tell me how to extract this train_all_val_all_bz_2_epoch_10_inject_init.pth.tar? I use tar -xf . However, I got error.
2. Can you tell me how to edit parser in the describe.py? I do not know what's the config_json mean. Do that mean I need to generate a json file for my image?

thanks,