Advances in Autonomous Driving and Depth Estimation

The field of autonomous driving and depth estimation is rapidly advancing with the development of novel deep learning frameworks and innovative applications. One of the key trends is the use of self-supervised learning methods, which eliminate the need for expensive sensors or large amounts of labeled data. Researchers are also exploring new architectures and techniques to improve the accuracy and efficiency of depth estimation and lane detection.

Notable progress has been made in monocular depth estimation, with new methods achieving state-of-the-art performance in low-light environments. Additionally, there is a growing interest in developing more efficient and scalable deep learning frameworks for autonomous driving, with a focus on reducing model size and inference time.

The development of new benchmarks and evaluation protocols is also an active area of research, with a focus on assessing the practical utility of depth foundation models in real-world applications.

Some particularly noteworthy papers include:

  • Depth3DLane, which proposes a novel dual-pathway framework for 3D lane detection that integrates self-supervised monocular depth estimation.
  • PRIX, which presents an efficient end-to-end driving architecture that operates using only camera data and achieves state-of-the-art performance on several benchmarks.
  • CHADET, which introduces a lightweight depth-completion network that generates accurate dense depth maps from RGB images and sparse depth points.
  • DepthDark, which proposes a robust foundation model for low-light monocular depth estimation that achieves state-of-the-art performance on several datasets.

Sources

Depth3DLane: Fusing Monocular 3D Lane Detection with Self-Supervised Monocular Depth Estimation

U-DREAM: Unsupervised Dereverberation guided by a Reverberation Model

Training Self-Supervised Depth Completion Using Sparse Measurements and a Single Image

Region-aware Depth Scale Adaptation with Sparse Measurements

Deep Learning Framework Testing via Heuristic Guidance Based on Multiple Model Measurements

CHADET: Cross-Hierarchical-Attention for Depth-Completion Using Unsupervised Lightweight Transformer

BenchDepth: Are We on the Right Way to Evaluate Depth Foundation Models?

PRIX: Learning to Plan from Raw Pixels for End-to-End Autonomous Driving

Reusing Attention for One-stage Lane Topology Understanding

DepthDark: Robust Monocular Depth Estimation for Low-Light Environments

Linear Memory SE(2) Invariant Attention

Built with on top of