OpenCV 4 with Python Blueprints - Second Edition

Using other detectors

In this chapter, we are using an object detector to get bounding boxes with their object types, which will be further processed by the Sort algorithm for tracking. In general, it does not matter by what exact means the boxes are obtained. In our case, we have used an SSD pre-trained model. Let's now understand how to replace it with a different model.

Let's first understand how we can use YOLO for this purpose. YOLO is also a single-stage detector and stands for You Only Look Once (YOLO). The original YOLO models are based on Darknet, which is another open-source neural network framework and is written in C++ and CUDA. OpenCV has the ability to load networks based on Darknet, similarly to how it loads TensorFlow models. ...