This paper provides a comprehensive survey of the application of Deep Learning (DL) in Internet of Things (IoT) data analytics, focusing on both big data and streaming data. It begins by defining IoT data characteristics and identifying two major treatments: IoT big data analytics and IoT streaming data analytics. The paper discusses why DL is a promising approach for these data types and applications, highlighting its potential and challenges. It reviews various DL architectures and algorithms, and analyzes major research attempts that have leveraged DL in IoT. The paper also discusses smart IoT devices that incorporate DL, as well as implementation approaches on fog and cloud centers. Finally, it outlines future research directions and challenges. The contributions of the paper include identifying key IoT data characteristics, reviewing state-of-the-art DL methods, and providing a guideline for using different DNNs in various IoT domains. The paper concludes with a summary of the main takeaways.This paper provides a comprehensive survey of the application of Deep Learning (DL) in Internet of Things (IoT) data analytics, focusing on both big data and streaming data. It begins by defining IoT data characteristics and identifying two major treatments: IoT big data analytics and IoT streaming data analytics. The paper discusses why DL is a promising approach for these data types and applications, highlighting its potential and challenges. It reviews various DL architectures and algorithms, and analyzes major research attempts that have leveraged DL in IoT. The paper also discusses smart IoT devices that incorporate DL, as well as implementation approaches on fog and cloud centers. Finally, it outlines future research directions and challenges. The contributions of the paper include identifying key IoT data characteristics, reviewing state-of-the-art DL methods, and providing a guideline for using different DNNs in various IoT domains. The paper concludes with a summary of the main takeaways.