The paper presents a novel method for detecting 3D model instances and estimating their 6D poses from RGB data in a single shot. The approach extends the SSD (Single Shot MultiBox Detector) paradigm to cover the full 6D pose space and is trained on synthetic model data. The method outperforms or competes with state-of-the-art methods that use RGB-D data on multiple challenging datasets while being significantly faster, achieving around 10Hz. The authors make their trained networks and detection code publicly available for reproducibility. The paper discusses the challenges in 3D detection and 6D pose estimation, highlighting the limitations of view-based methods and the advantages of color-based approaches. The methodology includes a detailed description of the network architecture, training process, and detection stage, including viewpoint scoring and pose refinement. Experimental results on multiple benchmark datasets demonstrate the effectiveness of the proposed method, showing competitive performance with RGB-D data-based methods and superior speed. The paper also addresses failure cases and suggests future directions for improving robustness to color deviations and loss term balancing.The paper presents a novel method for detecting 3D model instances and estimating their 6D poses from RGB data in a single shot. The approach extends the SSD (Single Shot MultiBox Detector) paradigm to cover the full 6D pose space and is trained on synthetic model data. The method outperforms or competes with state-of-the-art methods that use RGB-D data on multiple challenging datasets while being significantly faster, achieving around 10Hz. The authors make their trained networks and detection code publicly available for reproducibility. The paper discusses the challenges in 3D detection and 6D pose estimation, highlighting the limitations of view-based methods and the advantages of color-based approaches. The methodology includes a detailed description of the network architecture, training process, and detection stage, including viewpoint scoring and pose refinement. Experimental results on multiple benchmark datasets demonstrate the effectiveness of the proposed method, showing competitive performance with RGB-D data-based methods and superior speed. The paper also addresses failure cases and suggests future directions for improving robustness to color deviations and loss term balancing.