OpenShape: Scaling Up 3D Shape Representation Towards Open-World Understanding

TL;DR

OpenShape is a scalable multi-modal framework that learns joint text, image, and 3D shape representations, enabling open-world 3D recognition and interactions with state-of-the-art zero-shot performance.

Contribution

It introduces a scalable training approach with data filtering, network scaling, and a novel hard negative mining module for improved 3D shape understanding.

Findings

Achieves 46.8% zero-shot accuracy on Objaverse-LVIS benchmark

Outperforms previous methods with 85.3% accuracy on ModelNet40

Encodes diverse visual and semantic concepts for fine-grained interactions

Abstract

We introduce OpenShape, a method for learning multi-modal joint representations of text, image, and point clouds. We adopt the commonly used multi-modal contrastive learning framework for representation alignment, but with a specific focus on scaling up 3D representations to enable open-world 3D shape understanding. To achieve this, we scale up training data by ensembling multiple 3D datasets and propose several strategies to automatically filter and enrich noisy text descriptions. We also explore and compare strategies for scaling 3D backbone networks and introduce a novel hard negative mining module for more efficient training. We evaluate OpenShape on zero-shot 3D classification benchmarks and demonstrate its superior capabilities for open-world recognition. Specifically, OpenShape achieves a zero-shot accuracy of 46.8% on the 1,156-category Objaverse-LVIS benchmark, compared to less than 10% for existing methods. OpenShape also achieves an accuracy of 85.3% on ModelNet40, outperforming previous zero-shot baseline methods by 20% and performing on par with some fully-supervised methods. Furthermore, we show that our learned embeddings encode a wide range of visual and semantic concepts (e.g., subcategories, color, shape, style) and facilitate fine-grained text-3D and image-3D interactions. Due to their alignment with CLIP embeddings, our learned shape representations can also be integrated with off-the-shelf CLIP-based models for various applications, such as point cloud captioning and point cloud-conditioned image generation.