This paper presents a real-time 3D reconstruction system using voxel hashing for large-scale and fine-grained volumetric reconstruction. The system uses a memory and speed efficient spatial hashing scheme that compresses space and allows for real-time access and updates of implicit surface data without the need for a regular or hierarchical grid data structure. Surface data is stored densely only where measurements are observed, and data can be streamed efficiently in or out of the hash table, allowing for further scalability during sensor motion. The system supports interactive reconstructions of a variety of scenes, reconstructing both fine-grained details and large-scale environments. All parts of the pipeline, including depth map pre-processing, camera pose estimation, depth map fusion, and surface rendering, are performed at real-time rates on commodity graphics hardware. The system is compared to current state-of-the-art online systems, illustrating improved performance and reconstruction quality. The system uses a simple hashing scheme to compactly store, access, and update an implicit surface representation. The hash table sparsely and efficiently stores and updates TSDFs. The system pipeline includes a hash table data structure that stores sub-blocks containing SDFs, called voxel blocks. Each occupied entry in the hash table refers to an allocated voxel block. The system supports efficient compression of volumetric TSDFs, efficient fusion of new TSDF samples, removal and garbage collection of voxel blocks, lightweight bidirectional streaming of voxel blocks between host and GPU, and efficient extraction of isosurfaces for rendering and camera pose estimation. The system uses a GPU-accelerated hash table to manage allocation and retrieval of voxel blocks. The system supports real-time streaming between GPU and host, allowing unbounded reconstructions. The system is tested on a variety of indoor and outdoor scenes, demonstrating both scale and quality, and was all reconstructed well above the 30Hz frame rate of the Kinect. The system uses a voxel size of 4mm for some scenes and 10mm for others. The system is compared to other methods, showing improved performance and reconstruction quality. The system uses a total of 34MB for the hash table and all auxiliary buffers, allowing a hash table with 2^21 entries. The system is efficient in terms of speed, quality, and scalability. The system is able to handle large scenes with high resolution and quality, and is suitable for real-time applications. The system is efficient in terms of memory usage and performance, and is able to handle large scenes with high resolution and quality. The system is able to handle large scenes with high resolution and quality, and is suitable for real-time applications. The system is efficient in terms of memory usage and performance, and is able to handle large scenes with high resolution and quality.This paper presents a real-time 3D reconstruction system using voxel hashing for large-scale and fine-grained volumetric reconstruction. The system uses a memory and speed efficient spatial hashing scheme that compresses space and allows for real-time access and updates of implicit surface data without the need for a regular or hierarchical grid data structure. Surface data is stored densely only where measurements are observed, and data can be streamed efficiently in or out of the hash table, allowing for further scalability during sensor motion. The system supports interactive reconstructions of a variety of scenes, reconstructing both fine-grained details and large-scale environments. All parts of the pipeline, including depth map pre-processing, camera pose estimation, depth map fusion, and surface rendering, are performed at real-time rates on commodity graphics hardware. The system is compared to current state-of-the-art online systems, illustrating improved performance and reconstruction quality. The system uses a simple hashing scheme to compactly store, access, and update an implicit surface representation. The hash table sparsely and efficiently stores and updates TSDFs. The system pipeline includes a hash table data structure that stores sub-blocks containing SDFs, called voxel blocks. Each occupied entry in the hash table refers to an allocated voxel block. The system supports efficient compression of volumetric TSDFs, efficient fusion of new TSDF samples, removal and garbage collection of voxel blocks, lightweight bidirectional streaming of voxel blocks between host and GPU, and efficient extraction of isosurfaces for rendering and camera pose estimation. The system uses a GPU-accelerated hash table to manage allocation and retrieval of voxel blocks. The system supports real-time streaming between GPU and host, allowing unbounded reconstructions. The system is tested on a variety of indoor and outdoor scenes, demonstrating both scale and quality, and was all reconstructed well above the 30Hz frame rate of the Kinect. The system uses a voxel size of 4mm for some scenes and 10mm for others. The system is compared to other methods, showing improved performance and reconstruction quality. The system uses a total of 34MB for the hash table and all auxiliary buffers, allowing a hash table with 2^21 entries. The system is efficient in terms of speed, quality, and scalability. The system is able to handle large scenes with high resolution and quality, and is suitable for real-time applications. The system is efficient in terms of memory usage and performance, and is able to handle large scenes with high resolution and quality. The system is able to handle large scenes with high resolution and quality, and is suitable for real-time applications. The system is efficient in terms of memory usage and performance, and is able to handle large scenes with high resolution and quality.