FairMOT | FairMOT : On the Fairness of Detection | Computer Vision library

def postprocess(pred_dets, pred_embs, threshold = 0.5):
    tracker = JDETracker()
    online_targets_dict = tracker.update(pred_dets, pred_embs)
    online_tlwhs = defaultdict(list)
    online_scores = defaultdict(list)
    online_ids = defaultdict(  

crowdhuman
   |——————images
   |        └——————train
   |        └——————val
   └——————labels_with_ids
   |         └——————train(empty)
   |         └——————val(empty)
   └------annotation_train.odgt
   └------annotation_val.odgt

cd src
python gen_lab  

@article{zhang2020fair,
  title={FairMOT: On the Fairness of Detection and Re-Identification in Multiple Object Tracking},
  author={Zhang, Yifu and Wang, Chunyu and Wang, Xinggang and Zeng, Wenjun and Liu, Wenyu},
  journal={arXiv preprint arXiv:200  

def fit(
        self, lr: float = 1e-4, lr_step: str = "20,27", num_epochs: int = 30
    ) -> None:
        """
        The main training loop.

        Args:
            lr: learning rate for batch size 32
            lr_step: when to drop learn  

def predict(
        self,
        im_or_video_path: str,
        conf_thres: float = 0.6,
        track_buffer: int = 30,
        min_box_area: float = 200,
        frame_rate: int = 30,
    ) -> Dict[int, List[TrackingBbox]]:
        """
         

def _get_dataloader(self, im_or_video_path: str) -> DataLoader:
        """
        Create a dataloader from images or video in the given path.

        Args:
            im_or_video_path: path to a root directory of images, or single video or ima