improve efficiency

[cv-lab-x]

hi, thanks for your greak work, could you give some advices to improve the efficiency with depth estimation and fusion?
how to combine fusibile to your work ?
Looking forward to your reply, thanks. @whoiszzj

[whoiszzj]

Thank you for your question. We have yet to propose a method to improve efficiency that retains the same time cost as ACMM. However, the CUDA code contains many redundant computations, which may be simplified. The fusion process is a single-thread process, which costs a lot of time. A good way is to convert it to a multi-thread.
Regarding the fusibile method you mentioned, we can try to integrate it by modifying the data format. We have yet to do in-depth research on this, but we look forward to your further work.