ckpt trained on this detectron2 framework: checkpoint_state_dict['model'] Official COCO pretrained CenterNet ckpt: checkpoint

Modified focal loss. Exactly the same as CornerNet. Runs faster and costs a little bit more memory Arguments: pred (batch x c x h x w

focal loss from CornerNet

Start a new tracklet

Affine the coordinates. Args: coords (ndarray): floating point array of shape Nx2. Each row is (x, y).

Apply AffineTransform for the image(s). Args: img (ndarray): of shape HxW, HxWxC, or NxHxWxC. The array can be

Args: rotated_box (tuple): a tuple containing (cnt_x, cnt_y, w, h, angle), where cnt_x and cnt_y are the center c

:param tracks: list[STrack] :param detections: list[BaseTrack] :param metric: :return: cost_matrix np.ndarray

:param src: [E, F_x], where E is the number of edges. :param tgt: [E, F_x], where E is the number of edges. :param edge_at

Get input of each time's (T1 or T2) node feature and assigned gt ids :param node_feat: List[Tensor(#img * #topk_det, #feat_dim)] -

Args: inputs: feature matrix with shape (batch_size, feat_dim) targets: ground truth labels with shape (num_classes)