Hadamard Layer to Improve Semantic Segmentation

TL;DR

The paper introduces the Hadamard Layer, a parameter-free, computationally efficient addition that enhances semantic segmentation models by promoting distributed internal class encoding, leading to improved performance.

Contribution

It proposes the Hadamard Layer, a novel, parameter-free module that improves semantic segmentation accuracy without increasing model complexity.

Findings

Significant performance improvements in semantic segmentation models.

The Hadamard layer enforces distributed class encoding.

Training remains stable and fast with the new layer.

Abstract

The Hadamard Layer, a simple and computationally efficient way to improve results in semantic segmentation tasks, is presented. This layer has no free parameters that require to be trained. Therefore it does not increase the number of model parameters, and the extra computational cost is marginal. Experimental results show that the new Hadamard layer substantially improves the performance of the investigated models (variants of the Pix2Pix model). The performance's improvement can be explained by the Hadamard layer forcing the network to produce an internal encoding of the classes so that all bins are active. Therefore, the network computation is more distributed. In a sort that the Hadamard layer requires that to change the predicted class, it is necessary to modify $2^{k-1}$ bins, assuming $k$ bins in the encoding. A specific loss function allows a stable and fast training convergence.