Numerical Coordinate Regression with Convolutional Neural Networks

TL;DR

This paper introduces DSNT, a differentiable, parameter-free layer for neural networks that improves coordinate regression accuracy and spatial generalization, outperforming heatmap matching in pose estimation tasks.

Contribution

The paper proposes DSNT, a novel differentiable layer that enhances coordinate regression in CNNs, offering better accuracy and efficiency over existing heatmap matching methods.

Findings

DSNT outperforms heatmap matching in pose estimation accuracy.

DSNT maintains good spatial generalization with low-resolution heatmaps.

The method is compatible with various CNN architectures.

Abstract

We study deep learning approaches to inferring numerical coordinates for points of interest in an input image. Existing convolutional neural network-based solutions to this problem either take a heatmap matching approach or regress to coordinates with a fully connected output layer. Neither of these approaches is ideal, since the former is not entirely differentiable, and the latter lacks inherent spatial generalization. We propose our differentiable spatial to numerical transform (DSNT) to fill this gap. The DSNT layer adds no trainable parameters, is fully differentiable, and exhibits good spatial generalization. Unlike heatmap matching, DSNT works well with low heatmap resolutions, so it can be dropped in as an output layer for a wide range of existing fully convolutional architectures. Consequently, DSNT offers a better trade-off between inference speed and prediction accuracy compared to existing techniques. When used to replace the popular heatmap matching approach used in almost all state-of-the-art methods for pose estimation, DSNT gives better prediction accuracy for all model architectures tested.