Less is More: Efficient Point Cloud Reconstruction via Multi-Head Decoders

TL;DR

This paper introduces a multi-head decoder architecture for point cloud reconstruction that improves performance by leveraging redundancy and diversity, challenging the notion that deeper models are always better.

Contribution

We propose a novel multi-head decoder design that reconstructs point clouds from multiple independent outputs, demonstrating improved accuracy and generalization over traditional deep decoders.

Findings

Multi-head decoders outperform single-head baselines in key metrics.

Increasing decoder depth beyond a point causes overfitting and performance degradation.

Diversity in outputs enhances reconstruction quality and robustness.

Abstract

We challenge the common assumption that deeper decoder architectures always yield better performance in point cloud reconstruction. Our analysis reveals that, beyond a certain depth, increasing decoder complexity leads to overfitting and degraded generalization. Additionally, we propose a novel multi-head decoder architecture that exploits the inherent redundancy in point clouds by reconstructing complete shapes from multiple independent heads, each operating on a distinct subset of points. The final output is obtained by concatenating the predictions from all heads, enhancing both diversity and fidelity. Extensive experiments on ModelNet40 and ShapeNetPart demonstrate that our approach achieves consistent improvements across key metrics--including Chamfer Distance (CD), Hausdorff Distance (HD), Earth Mover's Distance (EMD), and F1-score--outperforming standard single-head baselines. Our findings highlight that output diversity and architectural design can be more critical than depth alone for effective and efficient point cloud reconstruction.