How to balance the size of the input image？

[Calsia]

the different scale of input images leads to different information of the scene. maybe one image includes just one object，for example one chair ， and the other image include the same chair and the environment around it. each image means different region . one is smaller than the other. and the first just in the second. how did this model balance it? or what did you do to fix it? thks for any reply~

[Calsia]

Besides, could you provide your G model, which generate the Panorama？ and，could you teach me how can i put the object in the exr scene. when i do it ，it just like fly in the sky

[Calsia]

by the way， i use the following file in the picture to test the light. and i choose/Fast Spatially-Varying/scene00_probe00/scene00_probe00.blend.

and i feed the regressionNetwork with /scene00_probe00/scene00_probe00.jpg. the output is below.

then i replace the environment texture in your blend file above with the predicted EL.

below is the rendered image by blender.

through the shadow direction, it seems like a little wrong. what could i do to fix it， i thought there will be a good result using your image. Sorry to bother you, but still hope you can reply me any time. thank you very very much

[Calsia]

Recently, I am also going to vote for related articles, so I hope you can cite your article and compare it with your results. So any reply is greatly encouraged. thank you again~

[fnzhan]

Besides, could you provide your G model, which generate the Panorama？ and，could you teach me how can i put the object in the exr scene. when i do it ，it just like fly in the sky

The project is conducted one year ago during the intern in Alibaba, the G model is saved in their server and I have no access to it now. But I have plan to re-train a G model (maybe next month), which should be quite simple. The 3D scenes in blender are designed by some designers in Alibaba.

[fnzhan]

There are two problems,

1. The 3D designed by the Alibaba designer should be rotated by 90 (or -90) in Z axis, I manually adjust it for my experiments.
2. The intensity is not corrected, which should be rectified by the generator. You can try to decrease the strength parameter.

[Calsia]

Thank you for your reply. So can I understand this way, I use the exr file obtained by rg to directly import blender, and then point the center of the image to the camera, and the correct result will be obtained？

[Calsia]

the different scale of input images leads to different information of the scene. maybe one image includes just one object，for example one chair ， and the other image include the same chair and the environment around it. each image means different region . one is smaller than the other. and the first just in the second. how did this model balance it? or what did you do to fix it? thks for any reply~

With these words, I would like to ask how to set the range of the input image to get better results？