search

cheind / py-motmetrics

:bar_chart: Benchmark multiple object trackers (MOT) in Python
MIT License
1.37k stars 258 forks source link

Everything gets 0 / NaN if confidence is -1 in motChallenge format #170

Open Sebastian-Vogt opened 2 years ago

Sebastian-Vogt commented 2 years ago

As stated in this issue #8 and in the comments following this one: All the results get 0s or NaN, if one insterts GT data with -1 as confidence. The mot challenge instructions state, that 0s should be ignored and any other value should be used. They also use -1 in their example.

jvlmdr commented 2 years ago

It indeed seems like the original matlab toolkit excludes only gt rows with confidence == 0. See here: https://github.com/dendorferpatrick/MOTChallengeEvalKit/blob/45098695e9c57999793181b658bebce380df6686/matlab_devkit/evaluateTracking.m#L50