ucuapps / top-view-multi-person-tracking Goto Github PK

Hello Ivan,

Thank you very much for sharing this dataset, I just wanted to ask about an explanation for the annotation.

An annotation file has 4 rows I assume that this means that this image has 4 people (please correct me if I'm wrong).
Each row has 5 numbers, the first number is always "0", so can you please just explain what are these numbers, and if they represent he bounding boxes, why are they fractions?

Final thing, if my assumption is correct and these numbers represent the bounding boxes and maybe the first number indicates for the class "which is always 0 because we are trying to track only people", can you explain where is the tracking ID for the people in different frames?

Thank you very much

Hello,

Thank you for creating the dataset.

Sorry, if this seems to be a silly question.

Usually, in the yolov5 format, if there are multiple instances of the same class, the bounding box and class information is present in different rows right ?
But in your dataset, suppose if there are two people in a frame, there is just one row in which there is a class information followed by bounding box co-ordinates for two instances. Can I use this for my object detection directly, if I want to use it to train my yolov8 ?
Or
Is there a way I can convert these into correct yolov5 format ?

Hi!
Can I get the code and weights of the trained model?

Thanks