Yuan Liu

Hi, I'm able to reproduce the results using the default setting. Maybe you can recheck if there is anything changed in the config files. Meanwhile, you may also reduce this...

I don't know the exact reason to cause this inconsistent results but you may try to reduce the parameter `freeze_inv_s_step` to 5000 or 0, which would improve the reconstruction quality...

The BRDF properties are pixel-wise defined on the vertex. You may also use UV mapping tools to extract a 2D UV map from the mesh and extract texture/roughness/metalness maps.

Hi, thanks for trying our codes. Maybe you need to change the name of the config. https://github.com/liuyuan-pal/NeRO/blob/3b4d421a646097e7d59557c5ea24f4281ab38ef1/configs/custom/kettle_material.yaml#L1 It will automatically load ckpt in `data/model/`.

Hi. The rendered images have RGBA channels. Then, I will convert the background to white.

Hi, the training is based on the pytorch_lightning and it is supposed to manage the resources correctly. You can see that the dataset class https://github.com/liuyuan-pal/SyncDreamer/blob/eb41a0c73748cbb028ac9b007b11f8be70d09e48/ldm/data/sync_dreamer.py#L57 which simply loads data here...

Hi, I'm also working on this problem. Directly scaling up consumes too much memory and seems to be infeasible here.

Hi, I'm not familiar with the runpod here so I'm afraid I'm not able to provide a runpod template here.

Hi, thank you and I'll upload them later.

Sorry for the late reply. The evaluation objects are here: https://connecthkuhk-my.sharepoint.com/:u:/g/personal/yuanly_connect_hku_hk/EUyeMurQcPxCtRy9MEvQevIBjjq8yu6LGOSqIAapZn9e4Q?e=pgmLCa