GPT4Point: A Unified Framework for Point-Language Understanding and Generation

TL;DR

GPT4Point tackles the scarcity and fragmentation of 3D language data by introducing a two-stage 3D MLLM that directly aligns 3D point clouds with language and supports controllable 3D generation. The Point-Q-Former fuses point and text features for downstream reasoning in an LLM and geometry-preserving diffusion-based 3D synthesis, while Pyramid-XL automates hierarchical point-text annotations from Objaverse-XL to enable large-scale training. A new Objaverse-LVIS benchmark evaluates 3D recognition, captioning, QA, and generation, and experiments show state-of-the-art zero-shot performance and meaningful controllable generation. Collectively, the approach advances direct 3D point-language modeling with scalable data and robust evaluation, offering significant implications for robotics, AR, and human–3D interactions.

Abstract

Multimodal Large Language Models (MLLMs) have excelled in 2D image-text comprehension and image generation, but their understanding of the 3D world is notably deficient, limiting progress in 3D language understanding and generation. To solve this problem, we introduce GPT4Point, an innovative groundbreaking point-language multimodal model designed specifically for unified 3D object understanding and generation within the MLLM framework. GPT4Point as a powerful 3D MLLM seamlessly can execute a variety of point-text reference tasks such as point-cloud captioning and Q&A. Additionally, GPT4Point is equipped with advanced capabilities for controllable 3D generation, it can get high-quality results through a low-quality point-text feature maintaining the geometric shapes and colors. To support the expansive needs of 3D object-text pairs, we develop Pyramid-XL, a point-language dataset annotation engine. It constructs a large-scale database over 1M objects of varied text granularity levels from the Objaverse-XL dataset, essential for training GPT4Point. A comprehensive benchmark has been proposed to evaluate 3D point-language understanding capabilities. In extensive evaluations, GPT4Point has demonstrated superior performance in understanding and generation.

Figures (17)

• Figure 1: Task examples of GPT4Point. It performs accurate 3D recognition, detailed captioning, precise Q&A, and high-quality controllable 3D generation. Additionally, GPT4Point excels in 3D anomalous object description, accurately assessing abnormal shapes like the multi-face object and the 3D generation failure case. It is a crucial ability in the assessment of generated 3D objects.
• Figure 2: The model architecture of GPT4Point for training. In Stage1, we employ a Bert-based Bert Point-Q-Former for point-text feature alignment through three point-text tasks. Then, in Stage2, an LLM is appended to train the model's text inference capabilities. A Point Cloud Diffusion is attached separately to train controlled text-to-3D generation which keeps the geometry shape and colors.
• Figure 3: Pyramid-XL: An automated point-text annotation engine. Directly inputting images into VLMs yields unsatisfactory results. We propose a progressive annotation approach with 3 levels of granularity, leveraging results from the previous level for precise outcomes.
• Figure 4: Examples of text inference using the GPT4Point with ViT-g and OPT6.7B after Instruct Finetuning. The table showcases its proficiency with point cloud input, excelling in tasks like detailed caption generation and point cloud-based question answering. This underscores our model's profound grasp of point cloud geometry and color, translating them into meaningful semantics.
• Figure 5: Object generated from Point-E fine-tuned on Cap3D Cap3D and our Pyramid-XL The first line shows Cap3D Cap3D fine-tuning results, while the second, using our Pyramid-XL Level 3 Dense Caption, outperforms Cap3D in geometry and color. This underscores the high quality of our text annotations.