The Deep-emotive v.1 is a CNN that recognizes emotions by the
human face’s pictures. In this context, the CNN’s structure creation
depends on several hyperparameters, which impact the results
positively or negatively. The Genetic Algorithm implementation
allows us to explore the search space of these hyperparameters
to find the best architecture for solving the problem. The defined
search space is formed by the combination of both the number
of convolutional layers and the fully connected ones, the number
of filters for each layer, the size of filters, the subsampling type,
and the number of nodes in the fully connected layer. This paper

proposes to improve the Deep-Emotive network with the imple-
mentation of Convolutional Neural Networks (CNNs) architectures

using Genetic Algorithms. The FER-2013 dataset was chosen to
classify seven emotions by images of facial expressions, as it had

the worst performance in the first version of the network, reach-
ing an accuracy of 60.71%. This dataset has images with common

problems for computer vision algorithms, such as occlusion, im-
balance, perspective, noises, as well as images that do not exist

in the context of emotions. The experiment’s results indicate that
the proposed approach can generate a CNN architecture with an
accuracy of 63,84% in the train set and 62,39% in the validation
set. Despite a low-performance rate, the experiments indicate that
the algorithm can generate more adapted individuals who have
already overcome the performance achieved by the first version of
the network defined empirically. Thus, results show potential for
exploitation in environments with more computational resources.