hkust-vgd/scenenn

How to generate instance segmentation?

Opened this issue · 5 comments

Dear SceneNN group,

I am a PhD student from Hebrew university, I meet some problem when i try to generate instance segmentation, could you give me a help.

I follow the instruction to download data from https://drive.google.com/drive/folders/0B-aa7y5Ox4eZWE8yMkRkNkU4Tk0 . But it seems like it requires additional processing to generate image and segmentation. And there is not instruction about how to do it. Could you give me a hint?

Thank you very much for your help.

Best Wishes,

Alex

Hi Alex,

To generate images from the dataset, you need to download the playback tool and the .oni files in the tools and oni folders, respectively. You can execute the playback tool as follows: Playback.exe 005.oni <output folder>.

For instance segmentation, we already provided 3D instance segmentation in the PLY files. You can take a look at the label field and the corresponding XML. For 2D instance segmentation, please take a look at our visualisation tool. You can also create your own 2D instance segmentation from 3D by back projection, given the camera poses.

Hope this helps.

Hi, Quang-Hieu,

Thank you for your instruction. I will try it out.

Best Wishes,

Alex

I use window, and run 'Playback.exe 700.oni .', it outputs:

Depth frames:0
Color frames:0

And I check the folders it creates (depth, image), it does not contain any image. I also try for 623.oni, it also produces empty folder. Could you tell me why this happen?

Hi, the annotation xml file (like 011.xml) contains , how to decode the value of obbox?