search

PerceivingSystems / bedlam_render

BEDLAM (CVPR 2023) render pipeline tools
https://bedlam.is.tuebingen.mpg.de/
Other
131 stars 6 forks source link

Seeking Guidance on Accessing Ground Truth Camera and SMPL Parameters #18

Closed niuzehai closed 10 months ago

niuzehai commented 10 months ago

I am currently trying to understand how to obtain the camera's intrinsic and extrinsic parameters, as well as the SMPL-X Ground Truth parameters for an image within UE 5.0.3. This information is crucial for my project, and I am seeking guidance on the best practices or methods to access these parameters.

Attempts and Research I have explored the documentation and various forums, but have not found a clear methodology or example that fits my use case.

Questions Are there built-in functions or plugins in UE 5.0.3 that allow for the extraction of camera intrinsic and extrinsic parameters? What is the recommended approach for accessing SMPL parameters for a given image?

Any guidance, references to documentation, or examples would be greatly appreciated.

Thank you in advance for your time and assistance.

tpsmpi commented 10 months ago

The ground truth data is logged in the ground_truth folder in .csv format. This assumes that you use the BE_GroundTruthLogger and follow the steps outlined in the Usage section at https://github.com/PerceivingSystems/bedlam_render/tree/main/unreal/render

Coordinate system for logged data is Unreal coordinate system.

To further process this data into a more ML friendly format see #4 and the related info in our BEDLAM ML repo

niuzehai commented 10 months ago

Dear tpsmpi,

Thank you very much for your prompt and detailed response. Your guidance has been instrumental in resolving my queries regarding the extraction of camera and SMPL-X Ground Truth parameters in Unreal Engine 5.0.3.

I followed your advice and explored the BE_GroundTruthLogger within the BEDLAM Render Control. The logging of ground truth data in the .csv format in the ground_truth folder, as mentioned in the Usage section at BEDLAM Render, was exactly what I needed.

Furthermore, I delved into the BEDLAM ML repo and found the functions for processing the data into a machine learning-friendly format to be very helpful. Specifically, the functions for converting camera intrinsic parameters (link), camera extrinsic parameters (link), as well as accessing the Beta (link) and Poses (link) parameters, have been particularly valuable.

Your assistance not only solved my immediate problem but also enhanced my understanding of handling such data in Unreal Engine for deep learning applications. I appreciate your time and effort in providing this crucial information.

Thank you once again for your invaluable help.