search

yuantianyuan01 / StreamMapNet

GNU General Public License v3.0
202 stars 17 forks source link

different performances between code and paper #16

Closed Michel-liu closed 11 months ago

Michel-liu commented 11 months ago

Hi~ Thank you for your great work! I have a question about the different performances between the code and paper. In Table 2, the AP of the new NuScenes split at 100x50m is 23.0, while the reported AP in the result of this code is 22.4. Besides, there are also many increasements than the paper. What is the reason for this phenomenon? Thanks!

yuantianyuan01 commented 11 months ago

Hi, thank you for your interest in our work. The results reported in this repo are produced by the released code and the ones in the paper are produced by previous dev version. The performance gap of all experiments are typically below 1 mAP which is minor and might due to some bug fixes.

Michel-liu commented 11 months ago

OK, I get it. Thank you again for your work and quick reply!