We get a negative result

cheind / py-motmetrics

:bar_chart: Benchmark multiple object trackers (MOT) in Python

MIT License

1.37k stars 258 forks source link

We get a negative result #103

Closed ChenSi521 closed 4 years ago

ChenSi521 commented 4 years ago

I used our validate dataset, but got a negative result. mota: -6.2% Can you tell me how to modify my gt.txt?

jvlmdr commented 4 years ago

While it is conventional to consider MOTA as a percentage, there is nothing inherently incorrect with a negative result for MOTA. Recall the definition MOTA = 1 - (FN + FP + IDSW) / GT. The number of false negatives and ID switches is bounded by the number of ground-truth boxes FN + IDSW <= GT. However, it is entirely possible for FP + (FN + IDSW) to exceed GT.

ChenSi521 commented 4 years ago

@jvlmdr thanks! I get it.

jvlmdr commented 4 years ago

Great!