Multi-Scale Occ: 4th Place Solution for CVPR 2023 3D Occupancy Prediction Challenge

TL;DR

This paper presents a multi-scale image feature-based method for 3D occupancy prediction, achieving 4th place in the CVPR 2023 challenge with a 49.36 mIoU score.

Contribution

It introduces Multi-Scale Occ, a simple yet effective approach leveraging multi-scale features and temporal fusion within the lift-splat-shoot framework for improved 3D occupancy prediction.

Findings

Achieved 4th place with 49.36 mIoU on the leaderboard.

Enhanced performance through model ensemble and test-time augmentation.

Utilized multi-scale features and temporal fusion for better 3D voxel representation.

Abstract

In this report, we present the 4th place solution for CVPR 2023 3D occupancy prediction challenge. We propose a simple method called Multi-Scale Occ for occupancy prediction based on lift-splat-shoot framework, which introduces multi-scale image features for generating better multi-scale 3D voxel features with temporal fusion of multiple past frames. Post-processing including model ensemble, test-time augmentation, and class-wise thresh are adopted to further boost the final performance. As shown on the leaderboard, our proposed occupancy prediction method ranks the 4th place with 49.36 mIoU.