# Semi-Supervised Single-View 3D Reconstruction via Prototype Shape Priors

Source: [Semi-Supervised Single-View 3D Reconstruction via Prototype Shape Priors](<https://www.sota2.com/research/paper/68fa6d841e209131cea3a494>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T00:36:22.217Z

Authors: Zhen Xing, Hengduo Li, Zuxuan Wu, Yu-Gang Jiang

Published: 2022-09-30T11:19:25.000Z

## Abstract

The performance of existing single-view 3D reconstruction methods heavily relies on large-scale 3D annotations. However, such annotations are tedious and expensive to collect. Semi-supervised learning serves as an alternative way to mitigate the need for manual labels, but remains unexplored in 3D reconstruction. Inspired by the recent success of semi-supervised image classification tasks, we propose SSP3D, a semi-supervised framework for 3D reconstruction. In particular, we introduce an attention-guided prototype shape prior module for guiding realistic object reconstruction. We further introduce a discriminator-guided module to incentivize better shape generation, as well as a regularizer to tolerate noisy training samples. On the ShapeNet benchmark, the proposed approach outperforms previous supervised methods by clear margins under various labeling ratios, (i.e., 1%, 5% , 10% and 20%). Moreover, our approach also performs well when transferring to real-world Pix3D datasets under labeling ratios of 10%. We also demonstrate our method could transfer to novel categories with few novel supervised data. Experiments on the popular ShapeNet dataset show that our method outperforms the zero-shot baseline by over 12% and we also perform rigorous ablations and analysis to validate our approach.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Single-view 3D Object Reconstruction on Pix3D 10% labeled data (S1)](<https://www.sota2.com/research/sota/single-view-3d-object-reconstruction-on-pix3d-10-percent-labeled-data-s1.md>): Single-view 3D Object Reconstruction; Pix3D 10% labeled data (S1); Chair Accuracy: 23.97.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2209.15383>)

[PDF](<https://arxiv.org/pdf/2209.15383.pdf>)
