A combination of object detection model with visual object tracking.
Trained to localize and track people, cars, bicycles, motorcycles, buses and trucks.
Inspired by UAVH
YOLOv3 with spatial pyramid pooling
trained on stanford dataset and VisDrone.
spp in Darknet:
model is taken from ultralytics
used opencv trackers
Spatial Pyramid Pooling paper https://ieeexplore.ieee.org/abstract/document/7005506
DCFCSR https://arxiv.org/pdf/1611.08461.pdf
weights can be found here