-
Notifications
You must be signed in to change notification settings - Fork 140
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
Some questions about model testing #45
Comments
Hi, Thank you for you interest in this project!
-Gaurav |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Dear Author,
Firstly, I would like to express my profound gratitude for your contributions and research in this field! I am currently working on multi-modal remote sensing for forest fire detection. Your recent introduction of the img2img method has inspired new directions in my research.
In remote sensing applications, acquiring registered multi-modal data (visible and thermal imaging) is exceedingly difficult. I was very intrigued by your project's handling of scenarios like Day to Night and Clear to Rainy transitions. I am interested in converting visible light images into registered thermal imaging to enrich my dataset with synthetic data.
To this end, I have trained a model locally using the training code you provided, utilizing some multi-modal forest fire data I have, which is already registered. I formatted the RGB and thermal images according to the structure used in your dataset.
The specific training command was as follows:
The model performed exceptionally well during validation, where the generated thermal images were almost identical to the target images.
10000 epoch:
![Clip_2024-05-10_10-33-23](https://private-user-images.githubusercontent.com/51520993/332326279-99571504-4dca-43af-9a05-a1e5d26331dc.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.0z6ss19d4WGPnMf4mjvnUasVFyqQbAMn_CpCUsAPfv4)
20000 epoch:
![Clip_2024-05-10_10-35-23](https://private-user-images.githubusercontent.com/51520993/332327039-ca29b072-0176-49a2-81a4-23c7882eaa7a.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.uMbs1z6Xgb_Qr1sOUX0T-QG7U9OEj3v082v7heFQvZc)
30026 epoch:
![Clip_2024-05-10_10-36-59](https://private-user-images.githubusercontent.com/51520993/332327280-273150df-ffd5-47d0-bc70-f604546f8c02.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MTk3MDYwMjAsIm5iZiI6MTcxOTcwNTcyMCwicGF0aCI6Ii81MTUyMDk5My8zMzIzMjcyODAtMjczMTUwZGYtZmZkNS00N2QwLWJjNzAtZjYwNDU0NmY4YzAyLnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNDA2MzAlMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjQwNjMwVDAwMDIwMFomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPWIzYjZmZjJjYzgwZWI2NzQwMzgwNzM5ZDQ3OTJhNjVmMzdkNzAxYTYzMzNhNGM4ZTc5MjgwMzU3NDQ3M2RmMGEmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0JmFjdG9yX2lkPTAma2V5X2lkPTAmcmVwb19pZD0wIn0.AEgZiGYlOJYVjfTQ6J0p0Z-fRNveTDBsc8DA7SE2dZo)
40001 epoch:
![Clip_2024-05-10_10-43-51](https://private-user-images.githubusercontent.com/51520993/332327282-ac7d1e50-1cad-4fb2-b997-dc3ab4cf0ff6.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.zy0NEleobngorXnCt7a2SZImzIp00Y9TyZb66t-ZXNY)
However, when I applied the model to test on other forest fire datasets, the results were disappointing and vastly different from what was expected.
The specific testing command was:
Given the above, I have several questions:
train_batch_size
to 1, which might have compromised effective learning.src/inference_paired.py
for inference with all parameters set to default. Do you think the distortion in inference could be related to the script's configuration?I look forward to your reply and thank you for taking the time to assist with these questions!
The text was updated successfully, but these errors were encountered: