thusiyuan / holistic_scene_parsing Goto Github PK

Hi,
I am trying to visualize 3D scenes, while inference.py only produces png images that includes depth and segmentation information. I am unable to figure out how osmesa is rendering scenes. I have compiled inference.py for more than an hour but the results are just getting optimized.
Please help. I want to visualize scene as a CAD model.

请问要修改代码来解决这个问题吗，我的系统是ubuntu18，在网上的关于ulimit的方案已经试过了，想问一下还有什么解决办法吗，谢谢。

Hi, thanks for sharing your great work.

I noticed that you need to detect the bounding boxes from input scene image and Retrieval Models from ShapeNet for each object. Is there the source code for Model Retrieving available?

Thanks again and looking forward to your response.

Hi, I encountered a problem of "Segmentation fault(Core Dump)" when I run "inference.py" with python 2.7.

Thank you for sharing the codebase.

Please complete the source code with trained models and other util files(if not being uploaded)

Regards.

Do you have any updates on this?

Hi,

Thank you for releasing the inference code.

The preprocessed images contain their respective pg files. Could you please release the script and instructions on how to do that for custom dataset images?

Thanks.

Hi,

First, thanks for sharing and congratulation for your paper.
I'm trying to replicate your results using your code and after running for 1h the inference using:
python inference.py -lo 11

I'm only getting purple or black png images of depth, segmentation and normal. (Also, I'm wondering where in the code you used a Neural Network to produce the 2D layout proposal.)

Thanks,
Elias