The capacity to learn, store, and represent knowledge are hallmarks of intelligent biological systems. In our pursuit to develop artificial intelligent systems, we have observed that while nature and computer science have converged on some approaches for learning, biological systems excel at quickly generalizing identification tasks with minimal supervision. Despite many advances in machine learning research, learning efficient knowledge representations in data and resource constrained environments is still a challenge. These observations and challenges motivate our work. We present OLMap, a biologically inspired, ensemble few-shot learning model. It uses modified fly olfactory circuits as an ensemble of weak learners and introduces few-shot supervision to generate generalized sparse binary representations that are suitable for image classification tasks. We tested the classification performance of OLMap using images from the public benchmark dataset STL-10. OLMap achieved promising and consistent results on image classification tasks with an average F1 score of \( 0.81 \) using different training datasets.