The aim of this bachelor thesis is to develop pose classification capabilities in an omnidirectional camera setup using transfer learning and open source models and datasets. The proposed system will be trained on a custom dataset of images captured in our Living Lab. The main challenge of this project is to adapt existing CNN architectures to work with omnidirectional images, which require special processing due to their distorted nature. To achieve our goals, we will first explore the state-of-the-art in pose classification using CNNs, with a focus on open source models and datasets. We will then use transfer learning to fine-tune existing CNNs on images captured in our Living Lab. The fine-tuned models will be evaluated on a test set, using standard metrics. Additionally, the resulting model will be applied for science communication purposes in the Living Lab, enhancing its value as a research and educational environment.