[CVPR 2022 Oral] Official repository for "MAXIM: Multi-Axis MLP for Image Processing". SOTA for denoising, deblurring, deraining, dehazing, and enhancement.
I'm converting the MAXIM models into Tensorflow.js models for use in Javascript.
When using fixed sizes, the model converts fine. However, using dynamic sizes (polymorphism) fails. For a discussion of why, see here.
This PR adjusts the UpSampleRatio block to avoid the casting to int, which allows dynamically sized inputs (although inputs must still be a multiple of 64, which must be enforced outside of the model call).
I'm converting the MAXIM models into Tensorflow.js models for use in Javascript.
When using fixed sizes, the model converts fine. However, using dynamic sizes (polymorphism) fails. For a discussion of why, see here.
This PR adjusts the
UpSampleRatio
block to avoid the casting toint
, which allows dynamically sized inputs (although inputs must still be a multiple of 64, which must be enforced outside of the model call).