Experimental Datasets and Processing Codes for the Semantic PHD Filter

The water bottle detection dataset and measurement model dataset for the paper titled "The Semantic PHD Filter for Multi-class Target Tracking: From Theory to Practice" by Jun Chen, Zhanteng Xie and Philip Dames, and the paper titled "Experimental Datasets and Processing Codes for the Semantic PHD Filter" by Zhanteng Xie, Jun Chen and Philip Dames

1. Detection dataset:  

Size: 
Total: 4870 images
Training: 4000 images
Validation: 870 images

Bottle Classes: Aquafina, Deer, Kirkland, Nestle

Format: PASCAL VOC, Darknet

Folder Structure:
 - Annotations: containing the xml label files in PASCAL VOC format
 - ImageSets: containing the training index files 
 - JPEGImages: containing the image data in jpg format
 - Labels: containing the txt label files in Darknet format

2. Measurement model dataset:

Format: ROSBAG

Duration: 19:59s (1199s)

Topics:
/darknet_ros/detection_image 3543 msgs : sensor_msgs/Image
/map 1 msg : nav_msgs/OccupancyGrid
/sphd_measurements 3585 msgs : sphd_msgs/SPHDMeasurements
/tf 142727 msgs : tf2_msgs/TFMessage
/tf_static 1 msg : tf2_msgs/TFMessage

Message Types:
nav_msgs/OccupancyGrid
sensor_msgs/Image
sphd_msgs/SPHDMeasurements
tf2_msgs/TFMessage

3. Processing codes:

Detection processing:
Zenodo: https://doi.org/10.5281/zenodo.7066045
GitHub: https://github.com/TempleRAIL/yolov3_bottle_detector

Measurement model processing:
Zenodo:  https://doi.org/10.5281/zenodo.7066050
GitHub: https://github.com/TempleRAIL/sphd_sensor_models