Indoor occupancy detection helps in creating energy efficient smart buildings which enable demand driven resource allocation. This research paper aims to enhance the occupancy detection models through application of transfer learning by leveraging knowledge from a pre trained source domain. The methodology involves PRAI-1581 dataset as source domain and fine-tuning it with BAE-6154 dataset as target domain collected from a second floor environment during both summer and winter seasons. This research addresses the challenge of limited target domain data by transferring learned features from the source domain, thereby reducing convergence time and computational resources while improving detection accuracy and precision. This work provides a comparison between the transfer learning model and the baseline model which is trained exclusively on the target dataset. The results indicate the efficacy of this approach in capturing seasonal variations in occupancy patterns across different environmental conditions. This research contributes to advancement of robust occupancy detection systems that can rapidly adapt to new environments.