chengzhag / deeppanocontext Goto Github PK

Well done for the excellent work, I got the predicted result after following the demo instructions, and I wander to get the 3D interactive result as https://chengzhag.github.io/publication/dpc/ shows, which tool can I use plz

It doesn't show any error. But while running project.py build command, it is getting stuck for long after giving the admin password.

ENV：conda env create -f environment.yaml

DATA: use Update data - Since iGibson has gone through a major update, their dataset download link has been updated. Please download the dataset here and follow the README to put the dataset into right places.

PROBLEM: Follow the README file. Run Training and Testing - First Stage - Save First Stage Outputs - CUDA_VISIBLE_DEVICES=0 WANDB_MODE=dryrun python main.py configs/first_stage_igibson.yaml --mode qtest --weight out/layout_estimation/<layout_estimation_id>/model_best.pth.
An error has occurred:
raise ValueError('all input arrays must have the same shape')
ValueError: all input arrays must have the same shape

How do I need to solve this problem?
Thank you.

Hi, I am trying to run DPC on google colab, but it is just not working, can you please give a notebook that is running DPC on colab.

When i run python project.py build i get above message.
I reinstalled libgl1-mesa-glx and checked whether it is installed by apt-get list --installed but still there's no libGL.so.1.2.0 in /usr/lib/x86_64-linux-gnu

how could i solve it?

Hello, i got problem when i run rendering.

message is like this.

I tried to change versions with export MESA_GL_VERSION_OVERRIDE, export MESA_GLSL_VERSION_OVERRIDE, export MESA_GLES_VERSION_OVERRIDE but it didn't work.

And i don't have 'shader' folder in '/root/anaconda3/envs/Pano3D/lib/python3.7/site-packages/gibson2/render/mesh_renderer'

My environment is docker container of remote server.

Could you help me?

This work is so excellent! I'm curious about how to obtain the results of total3d and im3d, as the layout they generate is a cuboid. Also, whether I can train total3d and im3d with current codes, or may you share their results? I want to learn how to combine several single-view results into a panoramic one.

The provided links in the readme contain model weights for the detector, layout estimation, and relation_scene_gcn modules, as well as the total3D weights.

Would it be possible to also provide the weights for the LIEN, LDIF and BEN modules?

Many thanks!

Could you please provide the processed training and test data(like Total3DUnderstanding) so that we can download it directly?

PermissionError: [Errno 13] Permission denied: '/home/anusree16/Dental/panorama diagnosis and 3D recon/DeepPanoContext-main/condaenv.syr0t84d.requirements.txt'

I looked into the code base and found that the 2D detector pipeline is unavailable.
I can see a comment that "Please follow Demo section to download weights for detector before we release full fine-tuning code for detector."

Is there any timeline for this?
If not is it possible for you to explain the steps if we want to upgrade it from detection or mask-rcnn to something else?
@chengzhag