This paper provides a comprehensive survey on the application of Deep Learning (DL) in the Internet of Things (IoT) domain, focusing on IoT big data analytics and streaming data analytics. The paper discusses the characteristics of IoT data, including large-scale streaming data, heterogeneity, time and space correlation, and high noise data. It also explores the potential of DL techniques for IoT data analytics, highlighting their advantages over traditional machine learning approaches, such as the ability to automatically extract features and improve accuracy. The paper reviews various DL architectures, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM), Autoencoders (AEs), Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), Restricted Boltzmann Machines (RBMs), and Deep Belief Networks (DBNs), and their applications in IoT. It also discusses the challenges and future research directions for DL in IoT, including the need for efficient DL models that can operate on resource-constrained devices and at the edge of the network. The paper emphasizes the importance of DL in enabling real-time and fast data analytics for IoT applications, which require quick decision-making and processing of data streams. It concludes with a summary of the main takeaways from the survey, highlighting the potential of DL in transforming IoT applications and improving their performance and efficiency.This paper provides a comprehensive survey on the application of Deep Learning (DL) in the Internet of Things (IoT) domain, focusing on IoT big data analytics and streaming data analytics. The paper discusses the characteristics of IoT data, including large-scale streaming data, heterogeneity, time and space correlation, and high noise data. It also explores the potential of DL techniques for IoT data analytics, highlighting their advantages over traditional machine learning approaches, such as the ability to automatically extract features and improve accuracy. The paper reviews various DL architectures, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM), Autoencoders (AEs), Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), Restricted Boltzmann Machines (RBMs), and Deep Belief Networks (DBNs), and their applications in IoT. It also discusses the challenges and future research directions for DL in IoT, including the need for efficient DL models that can operate on resource-constrained devices and at the edge of the network. The paper emphasizes the importance of DL in enabling real-time and fast data analytics for IoT applications, which require quick decision-making and processing of data streams. It concludes with a summary of the main takeaways from the survey, highlighting the potential of DL in transforming IoT applications and improving their performance and efficiency.