Simultaneous Localization And Mapping (SLAM) is the task of constructing a map of an unknown environment while simultaneously maintaining an estimate of the agent’s localization within this map. SLAM exists in many variants and is used in many different contexts. Dense 3D SLAM is one of the most important variants, which aims to create dense 3D maps that can be used for applications in fields such as civil engineering, security, virtual reality or autonomous robotics. The current standard method for dense 3D SLAM uses scanning Light Detection And Ranging (LiDAR) devices and Inertial Measurement Units (IMU), typically mounted on wheeled rovers for increased stability. As a whole, this setup allows a great mapping accuracy but is heavy, hard to navigate in non-flat environments, power-hungry and expensive. In this work, we aim to perform dense 3D SLAM with lightweight, portable sensors. More specifically, we use a specific type of camera that perceives 3D information with the same principle as LiDAR devices but in a miniaturized, less power-hungry manner. This allows the camera to be carried by hand or worn (on a helmet, for example) but comes at the cost of a decreased sensing accuracy. Moreover, carrying the camera by hand or wearing it on a helmet instead of setting it up on a wheeled rover makes it much less stable, making it even harder to estimate in the localization step of the problem. Therefore, most of our efforts go into the adaptation of existing LiDAR-inertial SLAM methods for noisier measurement with a more complex motion model. To do so, we develop a probabilistic SLAM framework based on the Error-State Iterative Kalman Filter (ESIKF) and implement various techniques to make the execution faster and less error-prone. With our experiments, we show that our method achieves competitive accuracy in different environments.