Building a 3D-model of an object from a set of its images using a neural network based on the NERF algorithm

The work was conducted as part of the development of a computer vision system for analyzing the environment, which could be utilized, for instance, by an autonomous mobile robot. This system involves using a camera to gather information about the surrounding environment. The paper presents methods for reconstructing three-dimensional models of objects solely from a set of 2D-images, using the NeRF algorithm to obtain a representation of a three-dimensional scene in the form of weights of a multilayer perceptron.Each method includes an estimate of the algorithm’s time consumption. Based on the data obtained, it was concluded that it is feasible to recognize the shapes of objects from a natural environment within 5-10 minutes, provided that the neural network training step is transferred to the server side.