# When and how CNNs generalize to out-of-distribution category-viewpoint combinations

arXiv: https://arxiv.org/abs/2007.08032 · HTML: https://mlsanity.com/paper/2007.08032/

## Key facts

- Cited by 22 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 15 Jul 2020.

## Built on by (most cited first, 22 of 22)

- [HuMMan: Multi-Modal 4D Human Dataset for Versatile Sensing and Modeling](https://mlsanity.com/paper/2204.13686/) (2022)
- [Playing for 3D Human Recovery](https://mlsanity.com/paper/2110.07588/) (2021)
- [Towards Viewpoint-Invariant Visual Recognition via Adversarial Training](https://mlsanity.com/paper/2307.10235/) (2023)
- [Rethinking Surgical Instrument Segmentation: A Background Image Can Be All You Need](https://mlsanity.com/paper/2206.11804/) (2022)
- [How Modular Should Neural Module Networks Be for Systematic Generalization?](https://mlsanity.com/paper/2106.08170/) (2021)
- [PSP-HDRI+: A Synthetic Dataset Generator for Pre-Training of Human-Centric Computer Vision Models](https://mlsanity.com/paper/2207.05025/) (2022)
- [The Robustness Limits of SoTA Vision Models to Natural Variation](https://mlsanity.com/paper/2210.13604/) (2022)
- [Does Progress On Object Recognition Benchmarks Improve Real-World Generalization?](https://mlsanity.com/paper/2307.13136/) (2023)
- [Towards Viewpoint Robustness in Bird's Eye View Segmentation](https://mlsanity.com/paper/2309.05192/) (2023)
- [Does Data Scaling Lead to Visual Compositional Generalization?](https://mlsanity.com/paper/2507.07102/) (2025)
- [Do Deep Networks Transfer Invariances Across Classes?](https://mlsanity.com/paper/2203.09739/) (2022)
- [Omniview-Tuning: Boosting Viewpoint Invariance of Vision-Language Pre-training Models](https://mlsanity.com/paper/2404.12139/) (2024)
- [Improving Viewpoint Robustness for Visual Recognition via Adversarial Training](https://mlsanity.com/paper/2307.11528/) (2023)
- [VR-Drive: Viewpoint-Robust End-to-End Driving with Feed-Forward 3D Gaussian Splatting](https://mlsanity.com/paper/2510.23205/) (2025)
- [WheelPose: Data Synthesis Techniques to Improve Pose Estimation Performance on Wheelchair Users](https://mlsanity.com/paper/2404.17063/) (2024)
- [AdvDreamer Unveils: Are Vision-Language Models Truly Ready for Real-World 3D Variations?](https://mlsanity.com/paper/2412.03002/) (2024)
- [Robust Self-Supervised Learning with Lie Groups](https://mlsanity.com/paper/2210.13356/) (2022)
- [Deep Crowd Anomaly Detection: State-of-the-Art, Challenges, and Future Research Directions](https://mlsanity.com/paper/2210.13927/) (2022)
- [Benchmarking Out-of-Distribution Generalization Capabilities of DNN-based Encoding Models for the Ventral Visual Cortex](https://mlsanity.com/paper/2406.16935/) (2024)
- [Three approaches to facilitate DNN generalization to objects in out-of-distribution orientations and illuminations](https://mlsanity.com/paper/2111.00131/) (2021)
- [Physically Disentangled Representations](https://mlsanity.com/paper/2204.05281/) (2022)
- [Automatic Assessment of Students' Classroom Engagement with Bias Mitigated Multi-task Model](https://mlsanity.com/paper/2510.22057/) (2025)

## Abstract (arXiv)

Object recognition and viewpoint estimation lie at the heart of visual understanding. Recent works suggest that convolutional neural networks (CNNs) fail to generalize to out-of-distribution (OOD) category-viewpoint combinations, ie. combinations not seen during training. In this paper, we investigate when and how such OOD generalization may be possible by evaluating CNNs trained to classify both object category and 3D viewpoint on OOD combinations, and identifying the neural mechanisms that facilitate such OOD generalization. We show that increasing the number of in-distribution combinations (ie. data diversity) substantially improves generalization to OOD combinations, even with the same amount of training data. We compare learning category and viewpoint in separate and shared network architectures, and observe starkly different trends on in-distribution and OOD combinations, ie. while shared networks are helpful in-distribution, separate networks significantly outperform shared ones at OOD combinations. Finally, we demonstrate that such OOD generalization is facilitated by the neural mechanism of specialization, ie. the emergence of two types of neurons -- neurons selective to category and invariant to viewpoint, and vice versa.