Obstacle detection is a critical research area for autonomous robots. It is especially important to detect small obstacles that can get entangled inside the robot. In addition, it can provide input to the movement and safety algorithm of autonomous devices by performing not only obstacle detection but also cliff detection. In this study, a tiny machine learning (TinyML) model that can run on a low-memory microcontroller and detect obstacles using multi-zone time-of-flight (ToF) sensors from STMicroelectronics is proposed. The proposed method ap-plied on an ST ARM based development kit. The object detection model achieved a high accuracy of over 90% on 5 different locations and obstacle presences.