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

only ~50% tampering mask detected #8

Open
Amit507017 opened this issue Sep 30, 2024 · 10 comments
Open

only ~50% tampering mask detected #8

Amit507017 opened this issue Sep 30, 2024 · 10 comments

Comments

@Amit507017
Copy link

Hello,

very nice work and thank you for sharing!

I followed the procedure for testing the approach exactly as mentioned in Readme. In the test, roughly in 50% of images only the tampered mask is predicted. In the remaining images complete white image is detected. Also when I ran the test multiple times, I see different set of images are correctly predicted. Any hints why this might be happening? Is someone else also experiencing this behaviour?

@xuanyuzhang21
Copy link
Owner

Could you share your testing steps and what is your testing data?

@Amit507017
Copy link
Author

I followed exactly same procedure as mentioned

Testing
Download the [testing dataset](https://drive.google.com/file/d/1s3HKFOzLokVplXV65Z6xcsBJ9qI91Qfv/view?usp=sharing) and place it in the "./dataset/valAGE-Set" and "./dataset/valAGE-Set-Mask". Download the pre-trained [checkpoint](https://drive.google.com/file/d/1w4e1gpdInAv7Lj_NQ7EGgmMuInyfUYgi/view?usp=sharing) and put it in the "./checkpoints".

cd code
python test.py -opt options/test_editguard.yml --ckpt ../checkpoints/clean.pth
To extract the tampered masks:

python maskextract.py --threshold 0.2

Also, the images for which correct tampering mask predicted are different when I tried running the script again.

@xuanyuzhang21
Copy link
Owner

Thanks for your reminder! Please set the ''degrade_shuffle=False'' in Line 25 of the "options/test_editguard.yml" since the uploaded checkpoint is trained under the clean condition. We have updated the file in our code~

@Amit507017
Copy link
Author

Thank you for checking. It is working quite well after that. Are you planning to release the model trained with degraded images as well?

@super-danny
Copy link

@Amit507017 , it's so good that you got the project running. How did you achieved this?
I currently have this working on a EC2 ubuntu instance and the project runs but I am unable to train it or test it currently.

Any hint on how you got it running, thanks

@user172894254
Copy link

@Amit507017 Hello, it's so good that you got the project running. I tried to reproduce this experiment, but I have some dependencies that cannot be downloaded, can you tell me what python environment you are configuring the environment in?

@user172894254
Copy link

@super-danny Hello, has your problem been solved? I can't get the project to run, I can't install some of the dependencies in it, can you install the dependencies normally?

@Amit507017
Copy link
Author

Sorry for late reply. I faced some problems with dependencies as well. Exactly I do not remember but I just relaxed some dependencies.

@qbl616686
Copy link

Sorry for late reply. I faced some problems with dependencies as well. Exactly I do not remember but I just relaxed some dependencies.

May I ask if you have finished debugging this project? May I ask if the code is complete? My second step python app.py doesn't work, and what's in Train2017.txt? Just the file name?

@SunSeaLucky
Copy link

Sorry for late reply. I faced some problems with dependencies as well. Exactly I do not remember but I just relaxed some dependencies.

May I ask if you have finished debugging this project? May I ask if the code is complete? My second step python app.py doesn't work, and what's in Train2017.txt? Just the file name?

After downloading COCO2017 and extracting all files, use ls train2017/ > train2017.txt to genertate the txt file.

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

6 participants