dos Santos, Claudio Filipi GoncalvesMoreira, Thierry Pinheiro [UNESP]Colombo, DaniloPapa, João Paulo [UNESP]2020-12-122020-12-122019-01-01Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), v. 11896 LNCS, p. 751-760.1611-33490302-9743http://hdl.handle.net/11449/201356Most Convolutional Neural Networks make use of subsampling layers to reduce dimensionality and keep only the most essential information, besides turning the model more robust to rotation and translation variations. One of the most common sampling methods is the one who keeps only the maximum value in a given region, known as max-pooling. In this study, we provide pieces of evidence that, by removing this subsampling layer and changing the stride of the convolution layer, one can obtain comparable results but much faster. Results on the gait recognition task show the robustness of the proposed approach, as well as its statistical similarity to other pooling methods.751-760engConvolutional Neural NetworksDeep learningGait recognitionTrabalho apresentado em evento10.1007/978-3-030-33904-3_712-s2.0-85075696640