Saved in:
Bibliographic Details
Main Author: Lengyel, Peter
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.06128
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908396520734720
author Lengyel, Peter
author_facet Lengyel, Peter
contents Predicting future states of dynamic agents is a fundamental task in autonomous driving. An expressive representation for this purpose is Occupancy Flow Fields, which provide a scalable and unified format for modeling motion, spatial extent, and multi-modal future distributions. While recent methods have achieved strong results using this representation, they often depend on high-quality vectorized inputs, which are unavailable or difficult to generate in practice, and the use of transformer-based architectures, which are computationally intensive and costly to deploy. To address these issues, we propose \textbf{Coupled Convolutional LSTM (CCLSTM)}, a lightweight, end-to-end trainable architecture based solely on convolutional operations. Without relying on vectorized inputs or self-attention mechanisms, CCLSTM effectively captures temporal dynamics and spatial occupancy-flow correlations using a compact recurrent convolutional structure. Despite its simplicity, CCLSTM achieves state-of-the-art performance on occupancy flow metrics and, as of this submission, ranks \(1^{\text{st}}\) in all metrics on the 2024 Waymo Occupancy and Flow Prediction Challenge leaderboard.
format Preprint
id arxiv_https___arxiv_org_abs_2506_06128
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CCLSTM: Coupled Convolutional Long-Short Term Memory Network for Occupancy Flow Forecasting
Lengyel, Peter
Computer Vision and Pattern Recognition
Predicting future states of dynamic agents is a fundamental task in autonomous driving. An expressive representation for this purpose is Occupancy Flow Fields, which provide a scalable and unified format for modeling motion, spatial extent, and multi-modal future distributions. While recent methods have achieved strong results using this representation, they often depend on high-quality vectorized inputs, which are unavailable or difficult to generate in practice, and the use of transformer-based architectures, which are computationally intensive and costly to deploy. To address these issues, we propose \textbf{Coupled Convolutional LSTM (CCLSTM)}, a lightweight, end-to-end trainable architecture based solely on convolutional operations. Without relying on vectorized inputs or self-attention mechanisms, CCLSTM effectively captures temporal dynamics and spatial occupancy-flow correlations using a compact recurrent convolutional structure. Despite its simplicity, CCLSTM achieves state-of-the-art performance on occupancy flow metrics and, as of this submission, ranks \(1^{\text{st}}\) in all metrics on the 2024 Waymo Occupancy and Flow Prediction Challenge leaderboard.
title CCLSTM: Coupled Convolutional Long-Short Term Memory Network for Occupancy Flow Forecasting
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.06128