Memory-Efficient Neural Network Deployment Methodology for Low-RAM Microcontrollers Using Quantization and Layer-Wise Model Partitioning