Data-to-Model Distillation: Data-Efficient Learning Framework