Open Paulkie99 opened 1 year ago
Hi @Paulkie99, the paper, written in 2022, was based on the 2022 challenge where GT detections were provided. This year, there has also been the 2023 challenge where GT detections weren't provided. Your result seems closer to the 2023 challenge baseline. Do you know which setup you ran?
Hi guys, I have followed the instructions for evaluating the ByteTrack benchmark on the test set, however I only get a HOTA of 43.627 instead of 47.225 as stated in your paper (w/o GT). Any idea why this might be? I want to make sure I understand how evaluation works before I start experimenting with my own models.