P3-SAM: Native 3D Part Segmentation
Changfeng Ma, Yang Li, Xinhao Yan, Jiachen Xu, Yunhan Yang, Chunshi Wang, Zibo Zhao, Yanwen Guo, Zhuo Chen, Chunchao Guo
2025-09-11
Summary
This paper introduces a new computer vision model, P3-SAM, that automatically breaks down 3D objects into their individual parts, like separating the wheel from the body of a car.
What's the problem?
Currently, it's difficult for computers to automatically and accurately identify the different parts that make up a complex 3D object. Existing methods struggle with complicated shapes and often require human help to finish the job, making it slow and inefficient.
What's the solution?
The researchers created P3-SAM, which is inspired by another image segmentation model called SAM. P3-SAM analyzes 3D data and uses a system of feature extraction, segmentation 'heads' (which identify potential parts), and a tool to predict how well those parts fit together. It also includes a way to automatically combine the identified parts into a complete breakdown of the object. They trained this model using a huge dataset of over 3.7 million 3D models.
Why it matters?
This research is important because it allows for better understanding of 3D shapes, makes it easier to reuse parts of 3D models, and opens up possibilities for things like automatically creating new 3D designs. It’s a step towards computers being able to 'see' and understand 3D objects as well as humans do.
Abstract
Segmenting 3D assets into their constituent parts is crucial for enhancing 3D understanding, facilitating model reuse, and supporting various applications such as part generation. However, current methods face limitations such as poor robustness when dealing with complex objects and cannot fully automate the process. In this paper, we propose a native 3D point-promptable part segmentation model termed P3-SAM, designed to fully automate the segmentation of any 3D objects into components. Inspired by SAM, P3-SAM consists of a feature extractor, multiple segmentation heads, and an IoU predictor, enabling interactive segmentation for users. We also propose an algorithm to automatically select and merge masks predicted by our model for part instance segmentation. Our model is trained on a newly built dataset containing nearly 3.7 million models with reasonable segmentation labels. Comparisons show that our method achieves precise segmentation results and strong robustness on any complex objects, attaining state-of-the-art performance. Our code will be released soon.