A simple Kalman filter for tracking bounding boxes in image space. The 8-dimensional state space x, y, a, h, vx, vy, va, vh co

Augmentation from CenterNet

Same as :class:`Checkpointer`, but is able to handle models in detectron & detectron2 model zoo, and apply conversions for legacy models.

Produce indices for inference. Inference needs to run on the __exact__ set of samples, therefore when the total number of samples is not

Affine Transform for CenterNet

The head used in CenterNet for object classification and box regression. It has three subnet, with a common structure but separate parameters

Affine Transformation from FairMOT official repository

nn.Module warpper for focal loss

currently, not loading from config file

The head used in CenterNet for object classification and box regression. It has three subnet, with a common structure but separate parameters

Generalized Tracker. Any models that contains the following two components: 1. Per-image object detector 2. Causal or online object track

Create a simple end-to-end predictor with the given config that runs on single device for a single input image. Compared to using the mo

Regression loss for an output tensor Arguments: output (batch x dim x h x w) mask (batch x max_objects) ind (batch x max_ob

Generalized Tracker. Any models that contains the following two components: 1. Per-image object detector 2. Causal or online object track

Triplet loss with hard positive/negative mining. Reference: Hermans et al. In Defense of the Triplet Loss for Person Re-Identification. arXiv: