model.jde

Description

🎯 Joint Detection and Embedding model for human detection and tracking.

class Node(config, **kwargs)[source]

Initializes and uses JDE tracking model to detect and track people from the supplied image frame.

JDE is a fast and high-performance multiple-object tracker that learns the object detection task and appearance embedding task simultaneously in a shared neural network.

Inputs

img (numpy.ndarray): A NumPy array of shape \((height, width, channels)\) containing the image data in BGR format.

Outputs

bboxes (numpy.ndarray): A NumPy array of shape \((N, 4)\) containing normalized bounding box coordinates of \(N\) detected objects. Each bounding box is represented as \((x_1, y_1, x_2, y_2)\) where \((x_1, y_1)\) is the top-left corner and \((x_2, y_2)\) is the bottom-right corner. The order corresponds to bbox_labels and bbox_scores.

bbox_labels (numpy.ndarray): A NumPy array of shape \((N)\) containing strings representing the labels of detected objects. The order corresponds to bboxes and bbox_scores.

bbox_scores (numpy.ndarray): A NumPy array of shape \((N)\) containing confidence scores \([0, 1]\) of detected objects. The order corresponds to bboxes and bbox_labels.

obj_attrs (Dict[str, Any]): A dictionary of attributes associated with each bounding box, in the same order as bboxes. Different nodes that produce this obj_attrs output type may contribute different attributes. model.fairmot produces the ids attribute which contains the tracking IDs of the detections.

Configs

  • weights_parent_dir (Optional[str]) – default = null.
    Change the parent directory where weights will be stored by replacing null with an absolute path to the desired directory.

  • iou_threshold (float) – default = 0.5.
    Threshold value for Intersecton-over-Union of detections.

  • nms_threshold (float) – default = 0.4.
    Threshold values for non-max suppression.

  • score_threshold (float) – default = 0.5.
    Object confidence score threshold.

  • min_box_area (int) – default = 200.
    Minimum value for area of detected bounding box. Calculated by \(width \times height\).

  • track_buffer (int) – default = 30.
    Threshold to remove track if track is lost for more frames than value.

References

Towards Real-Time Multi-Object Tracking: https://arxiv.org/abs/1909.12605v2

Model weights trained by: https://github.com/Zhongdao/Towards-Realtime-MOT