The present study shows an experimental deductive methodology based on neural networks for object recognition using CNN. Our objective is to generate a prototype which is based on a features map in combination with RPN and a trunk clipping proposal using TNET for 3D detection, given by object recognition models of the KITTI platform ,focused especially on AVOD and FPOINTNET, obtaining greater precision in smaller objects which are easily discarded by the point cloud provided by the 3D LIDAR HDL-64 laser sensor but not by the features mapping.
|Translated title of the contribution||Analysis of object detection algorithms for the creation of a prototype based on the fusion of two recognition models|
|Original language||Spanish (Ecuador)|
|Journal||ProSciences - Revista de Producción, Ciencias e Investigación|
|State||Published - 29 Mar 2019|