Seismic interpretation is a fundamental process for hydrocarbon exploration. This activity consists of identifying geological information through the processing and analysis of seismic data. With seismic data s rapid growth and complexity, manual seismic facies analysis has become a significant challenge. Mapping seismic facies is a time-consuming process that requires specialized professionals. The objective of this work is to apply multi-attribute classification using an encoder-decoder neural network to map the seismic facies and assist in the interpretation process. A set of seismic attributes were calculated using Opendtect version 6.6 software from the amplitude data contained in the Facies-Mark Dataset. These being: Energy, Pseudo Relief, Instant Phase, and Texture were all selected by an interpreter. The loss function used by the network was weighted categorical cross-entropy, because the classes are considerably unbalanced. The training was performed in the inlines and crosslines directions for the respective combinations: attributes, attribute + amplitude, and only the amplitude. The results based on the frequency weighted intersection over union (FWIU) metric showed that the attributes along with the amplitude obtained the best result, 85.73 percent, compared to the other combinations mentioned. In direct comparison with the work that inspired this dissertation, multi-attribute performed better.