Prediction of Occluded Pedestrians in Road Scenes using Human-like Reasoning: Insights from the OccluRoads Dataset

A. N. Melo, S. Martín, C. Salinas, M. A. Sotelo · 2024

Read Prediction of Occluded Pedestrians in Road Scenes using Human-like Reasoning: Insights from the OccluRoads Dataset by A. N. Melo, S. Martín, C. Salinas, M. A. Sotelo (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-06549.txt.

Description

Angie Nataly Melo Castillo and co-authors tackle hidden pedestrians in road scenes in a 2024 paper. Self-driving cars must spot people on foot, yet people partly or fully hidden behind objects remain hard to detect, even for strong current systems. Human drivers use scene context to guess where a hidden person may be, and the team aimed to copy that skill.

They built OccluRoads, a data set of real and virtual road scenes with hidden pedestrians, all labeled with rich context notes. Their pipeline joins a knowledge graph, graph embeddings, and Bayesian reasoning to predict whether a hidden person is present. It reached a strong score and beat standard machine learning models by up to 42 percent.

Catalog record

CREATORA. N. Melo, S. Martín, C. Salinas, M. A. Sotelo
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE27 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-06549

Browse the full library · About the Tower of Babel