Chained Predictions Using Convolutional Neural Networks

TL;DR

This paper introduces a convolutional neural network-based sequence-to-sequence model for structured output prediction in vision tasks, demonstrating improved performance in human pose estimation by predicting outputs sequentially with dependency on previous predictions.

Contribution

It adapts sequence-to-sequence models with CNNs for spatial localization, exploring weight sharing and demonstrating state-of-the-art results in human pose estimation.

Findings

Chained predictions outperform previous methods in human pose estimation.

Untied weights are effective for fixed-structure problems.

Sequential prediction improves spatial localization accuracy.

Abstract

In this paper, we present an adaptation of the sequence-to-sequence model for structured output prediction in vision tasks. In this model the output variables for a given input are predicted sequentially using neural networks. The prediction for each output variable depends not only on the input but also on the previously predicted output variables. The model is applied to spatial localization tasks and uses convolutional neural networks (CNNs) for processing input images and a multi-scale deconvolutional architecture for making spatial predictions at each time step. We explore the impact of weight sharing with a recurrent connection matrix between consecutive predictions, and compare it to a formulation where these weights are not tied. Untied weights are particularly suited for problems with a fixed sized structure, where different classes of output are predicted in different steps. We show that chained predictions achieve top performing results on human pose estimation from single images and videos.