Accelerate the entire BA process using the Ceres GPU version

[gentleman-zhong]

Thank you for your excellent work on this project!

I'm currently working with the incremental reconstruction pipeline and find that the overall process is significantly slower than expected. After profiling the runtime, I identified that over half of the total time is spent in the BA (Bundle Adjustment) stage.

To accelerate this process, I attempted to use the GPU version of Ceres Solver (version ≥ 2.1.0). I successfully compiled and ran the project with GPU-enabled Ceres:

• The compilation completed without errors
• GPU memory is occupied as expected during runtime
• However, the GPU utilization remains extremely low (close to 0% or single-digit percentages)

I initially suspected this was due to small problem size, but the low GPU utilization persisted even after I intentionally increased the scale of the BA problem (more points/frames).

I would greatly appreciate it if you could clarify:

1. Does the incremental reconstruction pipeline officially support GPU acceleration for the BA stage?
2. If GPU acceleration is supported, are there any specific configurations/modifications required to fully utilize GPU resources?
3. Are there any other recommended optimization strategies to speed up the incremental reconstruction process (either GPU-based or CPU-based)?

Thank you again for your time and contributions to this project!

[github-actions]

Could we move this conversation over to the forum at https://community.opendronemap.org? The forum is the right place to ask questions (we try to keep the GitHub issue tracker for feature requests and bugs only). Thank you!

p.s. I'm just an automated script, not a human being.

[YanNoun]

Hey ! You might want to use my fork, and better, check that it works correctly in ODM, see the PR : OpenDroneMap/OpenSfM#46

Cheers,

Yann