# Probing Fine-Grained Action Understanding and Cross-View Generalization of Foundation Models

arXiv: https://arxiv.org/abs/2407.15605 · HTML: https://mlsanity.com/paper/2407.15605/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 26 of its 33 references that are in the map.
- First posted on arXiv 22 Jul 2024.

## Builds on (most cited first, 26 of 26)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation](https://mlsanity.com/paper/2201.12086/) (2022)
- [On the Opportunities and Risks of Foundation Models](https://mlsanity.com/paper/2108.07258/) (2021)
- [Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset](https://mlsanity.com/paper/1705.07750/) (2017)
- [ViViT: A Video Vision Transformer](https://mlsanity.com/paper/2103.15691/) (2021)
- [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://mlsanity.com/paper/2203.12602/) (2022)
- [NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis](https://mlsanity.com/paper/1604.02808/) (2016)
- [CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retrieval](https://mlsanity.com/paper/2104.08860/) (2021)
- [NTU RGB+D 120: A Large-Scale Benchmark for 3D Human Activity Understanding](https://mlsanity.com/paper/1905.04757/) (2019)
- [Revisiting Feature Prediction for Learning Visual Representations from Video](https://mlsanity.com/paper/2404.08471/) (2024)
- [Learning Spatio-Temporal Representation with Pseudo-3D Residual Networks](https://mlsanity.com/paper/1711.10305/) (2017)
- [Channel-wise Topology Refinement Graph Convolution for Skeleton-Based Action Recognition](https://mlsanity.com/paper/2107.12213/) (2021)
- [Assembly101: A Large-Scale Multi-View Video Dataset for Understanding Procedural Activities](https://mlsanity.com/paper/2203.14712/) (2022)
- [X-CLIP: End-to-End Multi-grained Contrastive Learning for Video-Text Retrieval](https://mlsanity.com/paper/2207.07285/) (2022)
- [Revisiting Skeleton-based Action Recognition](https://mlsanity.com/paper/2104.13586/) (2021)
- [FineGym: A Hierarchical Video Dataset for Fine-grained Action Understanding](https://mlsanity.com/paper/2004.06704/) (2020)
- [Cross-view Action Modeling, Learning and Recognition](https://mlsanity.com/paper/1405.2941/) (2014)
- [The IKEA ASM Dataset: Understanding People Assembling Furniture through Actions, Objects and Pose](https://mlsanity.com/paper/2007.00394/) (2020)
- [CAT: Cross Attention in Vision Transformer](https://mlsanity.com/paper/2106.05786/) (2021)
- [TransDARC: Transformer-based Driver Activity Recognition with Latent Space Feature Calibration](https://mlsanity.com/paper/2203.00927/) (2022)
- [ViewCLR: Learning Self-supervised Video Representation for Unseen Viewpoints](https://mlsanity.com/paper/2112.03905/) (2021)
- [DVANet: Disentangling View and Action Features for Multi-View Action Recognition](https://mlsanity.com/paper/2312.05719/) (2023)
- [Coarse Temporal Attention Network (CTA-Net) for Driver's Activity Recognition](https://mlsanity.com/paper/2101.06636/) (2021)
- [How Object Information Improves Skeleton-based Human Action Recognition in Assembly Tasks](https://mlsanity.com/paper/2306.05844/) (2023)

## Built on by (most cited first, 2 of 2)

- [Hierarchical and Multimodal Data for Daily Activity Understanding](https://mlsanity.com/paper/2504.17696/) (2025)
- [Cross-View Cross-Modal Unsupervised Domain Adaptation for Driver Monitoring System](https://mlsanity.com/paper/2511.12196/) (2025)

## Abstract (arXiv)

Foundation models (FMs) are large neural networks trained on broad datasets, excelling in downstream tasks with minimal fine-tuning. Human activity recognition in video has advanced with FMs, driven by competition among different architectures. However, high accuracies on standard benchmarks can draw an artificially rosy picture, as they often overlook real-world factors like changing camera perspectives. Popular benchmarks, mostly from YouTube or movies, offer diverse views but only coarse actions, which are insufficient for use-cases needing fine-grained, domain-specific actions. Domain-specific datasets (e.g., for industrial assembly) typically use data from limited static perspectives. This paper empirically evaluates how perspective changes affect different FMs in fine-grained human activity recognition. We compare multiple backbone architectures and design choices, including image- and video- based models, and various strategies for temporal information fusion, including commonly used score averaging and more novel attention-based temporal aggregation mechanisms. This is the first systematic study of different foundation models and specific design choices for human activity recognition from unknown views, conducted with the goal to provide guidance for backbone- and temporal- fusion scheme selection. Code and models will be made publicly available to the community.