Using syntactic methods and LSTM to the recognition of objects visual patterns
Gilberto Astolfi, Vanessa Aparecida de Moares Weber, Adair da Silva Oliveira, Geazy Vilharva Menezes, Nícolas Alessandro de Souza Belete, Everton Castel�ão Tetila, Hemerson Pistori · 2019
In this paper, we have designed a new approach to represent and recognize objects visual patterns using syntactic methods. We capture relevant information from an object and associate them with symbols of an alphabet. After that, we derive a string from the object and in put it to LSTM. The idea is to train LSTM with objects visual patterns encapsulated in the strings. We conducted an experiment using soybean crops aerial images captured by an Unmanned Aerial Vehicle (UAV), and we reached an average F-measure of 91%.