POSTECH-CVLab/point-transformer

Quick question about performance

thangvubk opened this issue · 1 comments

Thank you very much for sharing your nice work. I have a quick question about the implementation. I have found that your original implementation did not reach the paper's reported results. Do you know what are the main reasons, which have been addressed in the author's provided code?

The original work may use some tricks to improve the performance. However, in fact, this Github repo may be the best implemented currently for point transformer, they give the result via s3dis area 5 . the result is close to the paper proposal. (miou 70% vs 70.4%)