Probabilistic Graph Attention Network with Conditional Kernels for Pixel-Wise Prediction

TL;DR

This paper introduces a probabilistic graph attention network with conditional kernels that improves multi-scale feature learning and fusion for pixel-wise prediction tasks, demonstrating superior results across multiple datasets and problems.

Contribution

It proposes a novel probabilistic graph attention network with Attention-Gated Conditional Random Fields and feature-dependent kernels for enhanced multi-scale feature fusion.

Findings

Outperforms existing methods on multiple datasets

Effective in diverse pixel-wise prediction tasks

Demonstrates improved structured feature learning

Abstract

Multi-scale representations deeply learned via convolutional neural networks have shown tremendous importance for various pixel-level prediction problems. In this paper we present a novel approach that advances the state of the art on pixel-level prediction in a fundamental aspect, i.e. structured multi-scale features learning and fusion. In contrast to previous works directly considering multi-scale feature maps obtained from the inner layers of a primary CNN architecture, and simply fusing the features with weighted averaging or concatenation, we propose a probabilistic graph attention network structure based on a novel Attention-Gated Conditional Random Fields (AG-CRFs) model for learning and fusing multi-scale representations in a principled manner. In order to further improve the learning capacity of the network structure, we propose to exploit feature dependant conditional kernels within the deep probabilistic framework. Extensive experiments are conducted on four publicly available datasets (i.e. BSDS500, NYUD-V2, KITTI, and Pascal-Context) and on three challenging pixel-wise prediction problems involving both discrete and continuous labels (i.e. monocular depth estimation, object contour prediction, and semantic segmentation). Quantitative and qualitative results demonstrate the effectiveness of the proposed latent AG-CRF model and the overall probabilistic graph attention network with feature conditional kernels for structured feature learning and pixel-wise prediction.