Deep Planar Parallax for Monocular Depth Estimation

TL;DR

This paper introduces PPNet, a novel monocular depth estimation method leveraging planar parallax geometry, flow-pretrain, and planar position embedding, achieving superior results on autonomous driving datasets.

Contribution

It proposes a new approach combining planar parallax geometry with flow-pretrain and PPE to improve depth estimation, especially for dynamic objects and slope variations.

Findings

PPNet outperforms existing methods on KITTI and WOD datasets.

Flow-pretrain enhances consecutive frame modeling.

Planar Position Embedding improves dynamic object handling.

Abstract

Recent research has highlighted the utility of Planar Parallax Geometry in monocular depth estimation. However, its potential has yet to be fully realized because networks rely heavily on appearance for depth prediction. Our in-depth analysis reveals that utilizing flow-pretrain can optimize the network's usage of consecutive frame modeling, leading to substantial performance enhancement. Additionally, we propose Planar Position Embedding (PPE) to handle dynamic objects that defy static scene assumptions and to tackle slope variations that are challenging to differentiate. Comprehensive experiments on autonomous driving datasets, namely KITTI and the Waymo Open Dataset (WOD), prove that our Planar Parallax Network (PPNet) significantly surpasses existing learning-based methods in performance.