# Quantifying and Learning Static vs. Dynamic Information in Deep Spatiotemporal Networks

arXiv: https://arxiv.org/abs/2211.01783 · HTML: https://mlsanity.com/paper/2211.01783/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 44 of its 80 references that are in the map.
- First posted on arXiv 3 Nov 2022.

## Builds on (most cited first, 30 of 44)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset](https://mlsanity.com/paper/1705.07750/) (2017)
- [Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation](https://mlsanity.com/paper/1802.02611/) (2018)
- [Non-local Neural Networks](https://mlsanity.com/paper/1711.07971/) (2017)
- [Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization](https://mlsanity.com/paper/1703.06868/) (2017)
- [SlowFast Networks for Video Recognition](https://mlsanity.com/paper/1812.03982/) (2018)
- [Is Space-Time Attention All You Need for Video Understanding?](https://mlsanity.com/paper/2102.05095/) (2021)
- [ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness](https://mlsanity.com/paper/1811.12231/) (2018)
- [The "something something" video database for learning and evaluating visual common sense](https://mlsanity.com/paper/1706.04261/) (2017)
- [Two-Stream Convolutional Networks for Action Recognition in Videos](https://mlsanity.com/paper/1406.2199/) (2014)
- [Multiscale Vision Transformers](https://mlsanity.com/paper/2104.11227/) (2021)
- [X3D: Expanding Architectures for Efficient Video Recognition](https://mlsanity.com/paper/2004.04730/) (2020)
- [End-to-End Video Instance Segmentation with Transformers](https://mlsanity.com/paper/2011.14503/) (2020)
- [Video Instance Segmentation](https://mlsanity.com/paper/1905.04804/) (2019)
- [YouTube-VOS: A Large-Scale Video Object Segmentation Benchmark](https://mlsanity.com/paper/1809.03327/) (2018)
- [Generic Attention-model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformers](https://mlsanity.com/paper/2103.15679/) (2021)
- [Keeping Your Eye on the Ball: Trajectory Attention in Video Transformers](https://mlsanity.com/paper/2106.05392/) (2021)
- [Learning Spatio-Temporal Features with 3D Residual Networks for Action Recognition](https://mlsanity.com/paper/1708.07632/) (2017)
- [REPAIR: Removing Representation Bias by Dataset Resampling](https://mlsanity.com/paper/1904.07911/) (2019)
- [Occluded Video Instance Segmentation: A Benchmark](https://mlsanity.com/paper/2102.01558/) (2021)
- [Self-supervised Video Object Segmentation by Motion Grouping](https://mlsanity.com/paper/2104.07658/) (2021)
- [A Survey on Deep Learning Technique for Video Segmentation](https://mlsanity.com/paper/2107.01153/) (2021)
- [Why Can't I Dance in the Mall? Learning to Mitigate Scene Bias in Action Recognition](https://mlsanity.com/paper/1912.05534/) (2019)
- [Something-Else: Compositional Action Recognition with Spatial-Temporal Interaction Networks](https://mlsanity.com/paper/1912.09930/) (2019)
- [A Comprehensive Study of Deep Video Action Recognition](https://mlsanity.com/paper/2012.06567/) (2020)
- [FusionSeg: Learning to combine motion and appearance for fully automatic segmention of generic objects in videos](https://mlsanity.com/paper/1701.05384/) (2017)
- [Only Time Can Tell: Discovering Temporal Data for Temporal Modeling](https://mlsanity.com/paper/1907.08340/) (2019)
- [Learning Video Object Segmentation with Visual Memory](https://mlsanity.com/paper/1704.05737/) (2017)
- [Semantic Video CNNs through Representation Warping](https://mlsanity.com/paper/1708.03088/) (2017)

## Built on by (most cited first, 7 of 7)

- [Universal Sparse Autoencoders: Interpretable Cross-Model Concept Alignment](https://mlsanity.com/paper/2502.03714/) (2025)
- [Understanding Video Transformers via Universal Concept Discovery](https://mlsanity.com/paper/2401.10831/) (2024)
- [Prisma: An Open Source Toolkit for Mechanistic Interpretability in Vision and Video](https://mlsanity.com/paper/2504.19475/) (2025)
- [HiERO: understanding the hierarchy of human behavior enhances reasoning on egocentric videos](https://mlsanity.com/paper/2505.12911/) (2025)
- [MODIX: A Training-Free Multimodal Information-Driven Positional Index Scaling for Vision-Language Models](https://mlsanity.com/paper/2604.12537/) (2026)
- [PixFoundation 2.0: Do Video Multi-Modal LLMs Use Motion in Visual Grounding?](https://mlsanity.com/paper/2509.02807/) (2025)
- [What, Where, and How: Probing Spatiotemporal Representations in Video Foundation Models](https://mlsanity.com/paper/2609.01551/) (2026)

## Abstract (arXiv)

There is limited understanding of the information captured by deep spatiotemporal models in their intermediate representations. For example, while evidence suggests that action recognition algorithms are heavily influenced by visual appearance in single frames, no quantitative methodology exists for evaluating such static bias in the latent representation compared to bias toward dynamics. We tackle this challenge by proposing an approach for quantifying the static and dynamic biases of any spatiotemporal model, and apply our approach to three tasks, action recognition, automatic video object segmentation (AVOS) and video instance segmentation (VIS). Our key findings are: (i) Most examined models are biased toward static information. (ii) Some datasets that are assumed to be biased toward dynamics are actually biased toward static information. (iii) Individual channels in an architecture can be biased toward static, dynamic or a combination of the two. (iv) Most models converge to their culminating biases in the first half of training. We then explore how these biases affect performance on dynamically biased datasets. For action recognition, we propose StaticDropout, a semantically guided dropout that debiases a model from static information toward dynamics. For AVOS, we design a better combination of fusion and cross connection layers compared with previous architectures.