This code implements the algorithm introduced in:

• Jinsong Zhang and Jean-François Lalonde, Learning High Dynamic Range from Outdoor Panoramas, International Conference on Computer Vision (ICCV), 2017.

This code takes as input a single LDR omnidirectional panorama, and converts it to HDR automatically.

For more details, please see our project webpage: http://www.jflalonde.ca/projects/learningHDR.

Important: if you use this code, please cite the paper above!

Execute the ldr2hdr.py to generate HDR image.

Download the data from our project webpage. The provided data contains all the LDR and HDR images used in the training and test. You may also need to download the SUN360 dataset to train the domain adaptation model. Example images can be found in the ./examples folder which have been aligned by centering the sun.

Our model is trained with TensorFlow; and OpenEXR is used to write HDR images.