The fields of computer vision and robotics are now more than ever being researched in tandem, with many aspects of both fields being developed in the context of the other. Within these two fields, the topic of 3D object reconstruction is a heavily researched one, using a series of 2D images in order to create 3D models. This area has a wide variety of possible applications in many different industries, ranging from tolerance checking within manufacturing, to geological surveys using aerial photography. In order to effectively automate these processes within the field of robotics, accurate benchmarking of the produced 3D models is needed in order to find optimal methods of object reconstruction.

The aim of this project is to create a pipeline system which allows for the easy implementation of different 3D object reconstruction methods and the measurement of the accuracy of the produced objects. Two different object reconstruction methods will be implemented, to which the user can add more, in order to test and compare different aspects of these methods. This pipeline should work within the Robot Operating System (ROS) and should have flexibility when dealing with different input and output parameters. The program should automatically assess the accuracy of given reconstruction methods when compared against the ground truth model. This project will be carried out in the simulation software gazebo.