gabrieleilertsen / hdrcnn

HDR image reconstruction from a single exposure using deep CNNs
https://computergraphics.on.liu.se/hdrcnn/
BSD 3-Clause "New" or "Revised" License
502 stars 101 forks source link

Regularization code for the existing code #54

Closed abhishri-medewar closed 1 year ago

abhishri-medewar commented 1 year ago

I have been working on implementing the regularization code by using the following document https://computergraphics.on.liu.se/temporally_stable_cnns/material/cvpr2019_supplementary.pdf as my reference. I am not able to understand which part of the training code(hdrcnn_train.py) should I add the following: Tfx = tf.contrib.image.transform(fx, T, interpolation = ’BILINEAR ’ ) ?? Any help would be appreciated.

gabrieleilertsen commented 1 year ago

Hi. Sorry for the late reply. Here is an example of integrating the regularization in the training code: hdrcnn_train.py.zip

On line 103-106 you can select what type of regularization, and which strength to use.

abhishri-medewar commented 1 year ago

Hi, Thank you for your reply. What data was used to verify the performance of the regularization method to check if it actually reduces the flicker issue?

gabrieleilertsen commented 1 year ago

Please have a look at the paper for specification of the datasets used. There were multiple experiments on colorization and single-image HDR reconstruction.