-
Notifications
You must be signed in to change notification settings - Fork 5
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
About the release date of the code and some questions #2
Comments
We are currently updating the Pytorch implementation here https://github.com/amberwangyili/neurop-pytorch.
Our main contribution, the neural color operator, is fundamentally different from CSRNet from aspects of both motivation (mimic color operators which have nice controllability with only one adjustable scalar) and design (preserve color operator properties through equivariant mapping so that it could be made simple and effective). Besides, reducing the 32D vector in CSRNet to a scalar value is also non-trivial. Naively replacing 32D vector with a 1D value in CSRNet would clearly lead to a drop in performance.
After fine-tuning the neurOps (combine three operators sequentially and train on the target dataset), the strength scalar is more like an exposed network parameter for us to further adjust the enhancement results. (please refer to the results we provided in the supplemental material) |
The text was updated successfully, but these errors were encountered: