Implementing Convolutional Neural Networks (CNNs) for computer vision tasks within Internet of Things (IoT) sensor nodes presents significant challenges due to the stringent computational, memory, and latency constraints inherent in these environments. To overcome these hurdles, researchers have employed various optimization techniques such as quantization, pruning, and model partitioning. While model partitioning alleviates the computational load on individual nodes, it does not reduce the overall system's computational requirements and introduces additional communication energy costs. Our research aims to conduct a comprehensive analysis of the impact of partitioning, quantization, and pruning on CNN models. We will investigate how these techniques affect the models' mean average precision and data size, crucial metrics for evaluating performance and efficiency in constrained environments. Given the need to train multiple CNN models extensively as part of our study, access to more substantial and faster computing resources is imperative. This access will enable us to accelerate our research, ensuring smoother and more efficient progress. Therefore, we are applying for the NAISS small compute resource to support this endeavor.