Many science-education tools lag behind current computer-vision practice and offer limited interactivity. We introduce a simple, lightweight, and modular teaching aid for railway track detection that integrates low-cost hardware with on-device vision models to enable intuitive visualization and multi-dimensional interaction. The research includes: (1) Design of a compact electromechanical platform by integrating a detection vehicle and track modules with a power unit, enabling fast assembly and part replacement; (2) Implementation of a recognition algorithm based on the Maix-I development board, where a streamlined program for visual track detection was developed. At the algorithmic level, we fine-tuned a MobileNet_0.75 model and established a complete pipeline of training, evaluation, and local deployment through the nncase toolchain. This approach demonstrates extensibility across both software and hardware platforms. The resulting educational tool has been validated through multiple outreach activities and has received widespread acclaim from educators.