In the first part of this chapter, we briefly described different IoT applications and their image detection-based decision-making. In addition, we briefly discussed two use cases: image detection-based road fault detection, and image detection-based solid waste sorting. The first application can detect potholes on the road using a smartphone camera or a Raspberry Pi camera. The second application detects different types of solid waste and sorts them according to smart recycling.
In the second part of the chapter, we briefly discussed transfer learning with a few example networks, and examined its usefulness in resource-constrained IoT applications. In addition, we discussed the rationale behind selecting a CNN, including two popular implementations, namely Inception V3 and Mobilenet V1. The rest of the chapter described all the necessary components of the DL pipeline...
