We present a method for combining Scale-Invariant Feature Transformation (SIFT) and Segment Anything Model (SAM) for the purpose of Video Object Segmentation (VOS). We combine the high accuracy constituted by SAM with a feature matching algorithm. SAM is limited to segmenting single images, as it has no module for exchanging information between frames. To enable SAM’s usage in a VOS context, we utilize SIFT as a feature matching algorithm, thus transferring positional information from one image to another. Our method is comparatively simple, which might cause problems to arise; however, this enabled us to vary and adapt the method more easily, which is also an advantage for future work. We discuss this pipeline’s capabilities, limitations and potential for improvement. This is also compared to similar me for implementing SAM. We validate our approach in an extensive qualitative study in the context of traffic. Further, we evaluate the real world usecase of 3-dimensional object tracking based on VOS data. Our approach is limited by the capabilities of SIFT; however, they were identical with the limitations of triangulation used in the task of object tracking, as this is also based on SIFT. Overall our approach does not achieve the robustness of similar models; however, its adaptability enables us to improve and specialize our model in future work.