Eckert comments

Results 11 comments of


                                            Eckert

Camera poses problem？cam data problem?

@a123935132 How did you solve this problem? I meet a similar problem. How do you generate the camera file? Thanks for your time.

run error

After install opencv3.4.5, run and got error: `./JointTG: error while loading shared libraries: libJointTexGeo.so.1: cannot open shared object file: No such file or directory` Is there somethings that I missed?

run error

Well, I solve it by setting `export LD_LIBRARY_PATH=/.../JointTG/lib:$LD_LIBRARY_PATH`. In addition, there are also need to install `libomp5`: `apt-get install -y libomp5 libomp-dev freeglut3 freeglut3-dev libxi-dev libxmu-dev mesa-utils libgl1-mesa-glx pkg-config libxau-dev...

Hello and thank you for your wonderful work. I would like to ask again if you have done some preprocessing on the scale of the depth map inferred by the depth estimation model, because I found that it is often difficult to directly use the depth map obtained by depth estimation for warp to align the target perspective pose and depth value. The scale size greatly affects the warp operation.

yes, we need to pre-process the depth map estimated by the depth estimation model, please refer the line 200 of [dataLoader](https://github.com/eckertzhang/Text2NeRF/blob/73c2745744fb26276aa7293fd9b7c4e9933977ae/dataLoader/scene_gen.py#L200) for more details.

360 degree scene

Sorry for the inconvenience. Please change the 'pose_traj=circle0' to 'pose_traj=circle'. Here, 'circle0' was used for early code testing and has been discarded.

no such file './Weights/econ_weights/ckpt/normal.ckpt'

You can follow ECON's instruction to download associated weights of ECON. https://github.com/YuliangXiu/ECON/blob/master/fetch_data.sh

Question about 'WEIGHT_PATH'

os.environ['WEIGHT_PATH'] should be updated as the weight path of pre-trained models, like stable diffusion and ECON.

Question about 'WEIGHT_PATH'

The pre-trained model comes from hugging face ('https://huggingface.co/runwayml/stable-diffusion-v1-5’) and ECON ('https://github.com/YuliangXiu/ECON/blob/master/fetch_data.sh')

Question about 'TORCH_HOME'

'TORCH_HOME' could be a local path to save weights of pre-trained models provided by pytorch. Actually, the default path is "/username/.cache/torch/hub/checkpoints".

Missing packages

‘contextual-loss-pytorch’ will not be used in the training/optimization stage. 'clip' can be download by: `pip install git+https://github.com/openai/CLIP.git`