Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

lack of files #2

Open
z972778371 opened this issue Mar 16, 2022 · 4 comments
Open

lack of files #2

z972778371 opened this issue Mar 16, 2022 · 4 comments

Comments

@z972778371
Copy link

I found there is no COCO14-train.p in my ./cider/data files.
There's problem to import resource in Main.py.
Could you share the pretrained models to run the Main.py in sample mode.
By the way, how to get the visualization pics like yours?

@zyj0021200
Copy link
Owner

zyj0021200 commented Mar 17, 2022 via email

@z972778371
Copy link
Author

z972778371 commented Mar 18, 2022 via email

@zyj0021200
Copy link
Owner

Hello and I have recently tested and updated my project. And there are a few things to take note.
1.I don't recommend running this project in Windows environment not because of the path problems I mentioned before but the file encoding and decoding thing. I encountered with the decoding error when performing METEOR calculation like 'UnicodeEncodeError: 'gbk' codec can't encode character '\xa0' in position 57: illegal multibyte sequence', this might be solvable for you but again I don't really have enough time to figure out; The "resource" package imported in Main.py is not needed in windows. So you could just ignored this. Also the code 'csv.field_size_limit(sys.maxsize)' in 'PreProcess/Generate_coco14_bottom_up_features_data.py' should be modified by exchanging a suitable large integer for 'sys.maxsize';
2.The COCO14-train.p is created by running 'PreProcess/CIDEr_idf_preproccess.py'. This file is used to perform SCST training for quick calculation of cider scores;
3.The visualization pictures is generated by the visualize_att and visualize_att_bboxes functions in 'Utils.py' , simply putting attention-weight tensors over the tensors of different image regions and using matplotlib to draw the pics;
Finally since I have graduated and not engaged in relative industries, I don't have enough computing resources right now to re-run the experiments but I think the performances are reproducible. Maybe later I will try to give out the pretrained models. But I also give a file 'train_NIC_COCO14.sh' to train NIC model and you can follow the parameter settings.

@z972778371
Copy link
Author

z972778371 commented Mar 30, 2022 via email

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants