First, thank you very much for your contribution- a StrongSORT fork has worked very well for my detection and tracking project!
I am tracking wildlife in videos. Sometimes, a bird is static across thousands of frames; I observe a bounding box when the bird is detected, and this box blinks on and off, which is not surprising. I have code to backfill the false negative frames.
I notice that even between very, very short non-detections (< 10 seconds), a static bird will be assigned a new tracking ID. My initial solution was to increase max_age a lot to give the model more leeway to "remember" that bird X is at position Y. However, no matter how high I push max_age, the model does not change its number of newly assigned tracking IDs.
First, thank you very much for your contribution- a StrongSORT fork has worked very well for my detection and tracking project!
I am tracking wildlife in videos. Sometimes, a bird is static across thousands of frames; I observe a bounding box when the bird is detected, and this box blinks on and off, which is not surprising. I have code to backfill the false negative frames.
I notice that even between very, very short non-detections (< 10 seconds), a static bird will be assigned a new tracking ID. My initial solution was to increase
max_age
a lot to give the model more leeway to "remember" that bird X is at position Y. However, no matter how high I pushmax_age
, the model does not change its number of newly assigned tracking IDs.Is there any solution to this problem? Thank you.