Download PDFOpen PDF in browser

Estimation of Height of a Shape a 2D Image from Its Shadow Using Neural Networks

EasyChair Preprint 10667

8 pagesDate: August 4, 2023

Abstract

Determining the height of objects in a 2D image from their shadow and developing a model capable of estimating the height of uniform geometric shapes and objects of different sizes from the shadow they cast, is the fundamental part of this research, which proposes the use of convolutional neural networks (CNN) as a Machine Learning (ML) technique for pattern detection, feature extraction from shadows and shapes in images. For this purpose, a dataset was constructed with photographic images of shapes or objects, as well as their shadows cast from different angles and locations. The structure of the proposed dataset is characterized by the name of the shape, the name of the shadow, the length of the shadow, the height of the shape and the angle of the light source, which together allow to improve the accuracy of the model. In this sense, the research focuses on the analysis of lights and shadows of different geometric shapes or objects within a 2D image, where the projected shadow is the information to be used and with which it is intended to determine the height of the shapes or objects. The main reason for this research is oriented towards people who have visual impairment either total or partial, and that from the touch pretend to define or differentiate an object, becoming impossible if asked to indicate which of these are present in a 2D image (photograph or painting). Therefore, it is essential that from a 2D image is to highlight the objects to be represented in a 2.5 model, where the height of the selected objects will be the key to create the model. In this sense the project to be developed seeks to determine the height of predefined objects in a two-dimensional image from its shadow.

Keyphrases: 2D, Convolutional Neural Network, Shadows, dataset, height, shapes

BibTeX entry
BibTeX does not have the right entry for preprints. This is a hack for producing the correct reference:
@booklet{EasyChair:10667,
  author    = {Julian Rene Muñoz Burbano and Pablo Emilio Jojoa Gómez and Fausto Miguel Castro Caicedo},
  title     = {Estimation of Height of a Shape a 2D Image from Its Shadow Using Neural Networks},
  howpublished = {EasyChair Preprint 10667},
  year      = {EasyChair, 2023}}
Download PDFOpen PDF in browser