Alex Bewley

This is true only if you have replaced detections with the ground truth. On Fri, 26 Feb. 2021, 16:25 R0flcopt3r, wrote: > Is it safe to assume that unmatched_dets are...

Prediction is called on all tracks matched and unmatched and I believe the track management you are referring to is on lines 248 to 250. Where if a track goes...

You will need to look closely to the output of the linear assignment. If the input is a new object you will have to wait until it is fully initialized...

You get the dets from an external detector. I provide a link to the one I used in the README. The code performs tracking (not detection) by associating the detections...

The latest version replaced a for loop with implicit broadcasting in numpy. This significantly improved speed making it slightly faster than using numba and with fewer dependencies. Hence numba is...

The latent state of the Kalman filter also estimates the velocity of the object. This can be thought of as a delta applied at each time-step to update the position...

Try removing the @jit decorator, then you shouldn't need numba. On Thu, 23 Apr. 2020, 19:31 GOBish, wrote: > I have been trying to implement SORT on my Raspberry pi...

Maybe you can provide more information on what the arguments represent. SORT takes one argument which is an array of boxes where each column is: left, top, right, bottom, score....

Can you please provide more details? E.g. your OS? which version of python? error message? You can also try adding the following to the top of your sort.py file: ```...

You are correct that this tracker (like all tracking by detection methods) are very sensitive to the input detections. However, to retain the identities for longers you can increase the...