In this chapter, we explored how a developer could get started with object detection applications using the OpenMV camera. We examined the machine learning technologies that drive this capability under the hood, such as CMSIS-NN. While training cannot be done on the target device, the inference can be executed on a resource-constrained processor.
Depending on the end application and the object that needs to be detected, a developer may be able to leverage existing datasets to train their model. Worst case scenario, a developer may need to acquire and classify the data themselves. With the knowledge gained in this chapter, you should now be able to train your own custom models and deploy them on the OpenMV camera. You can also leverage the existing, pretrained models and examples to develop extremely sophisticated applications.
In the next chapter, we are going to discuss...
