Determination of 3D Coordinates of Objects from Image with Deep Learning Model

We can capture an image or record a video by using a camera, but these frames take the two-dimensional representation of the objects. The main drawback of two-dimensional representation is the depth or perspective view between objects is unknown, sometimes an object is overlapped with another and could not be captured. This drawback can be solved if we could represent the objects by giving them three-dimensional coordinates. This paper proposes a lightweight algorithm that can distinguish different objects and can provide a three-dimensional coordinate to an object. This solution is developed using You Only Look Once (YOLO) for object detection and classification with the COCO dataset, and two cameras for observing the perspective between the objects. The system can provide portability and is easy to use. The algorithm can be used with different types of deep learning models.