Repository navigation
Any chance to improve the tracking evaluation code? #884
Description
Activity
@Kemo-Huang thanks for your suggestion!
If would be great if you could provide some profiling of the evaluation code in order to quantify the "slow" behavior you observe
In addition, do also provide examples of the methods in
motmetrics>1.1.3which are faster than their counterparts inmotmetrics<=1.1.3(and are used innuscenes-devkit)
#300 uses motmetrics <= 1.1.3. However, this version of motmetrics runs very slow. I wonder whether the evaluation code can be adapted to the latest version in the future.
Moreover, the pandas warning is too annoying:
nuscenes\eval\tracking\mot.py:125: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead. r = r.append(copy)Really wish the developers could make improvement for these.