Camera Movement Classification in Historical Footage: A Comparative Study of Deep Video Models

2025-10-16

AI Summary

This work presents the first systematic evaluation of modern deep video camera motion classification (CMC) models on historical archival film—characterized by low quality, degradation, and aging artifacts. Addressing the gap in prior research, which has largely overlooked historical footage, we benchmark five state-of-the-art models—including spatiotemporal Transformers such as Video Swin Transformer—on the expert-annotated HISTORIAN dataset. Results show Video Swin Transformer achieves the highest accuracy (80.25%), demonstrating robust adaptability to low-quality, small-sample video. However, all models exhibit consistent performance degradation under noise, low resolution, and reduced frame rates. Our core contributions are: (1) establishing the first standardized benchmark framework for CMC on historical video; and (2) identifying concrete improvement directions—specifically, multimodal input fusion and explicit temporal modeling—to advance reproducible, evidence-based archival video content understanding.

Technology Category

• Computer Vision: Video Understanding & Activity Analysis
• Machine Learning: Deep Neural Architectures and Foundation Models
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Camera movement conveys spatial and narrative information essential for understanding video content. While recent camera movement classification (CMC) methods perform well on modern datasets, their generalization to historical footage remains unexplored. This paper presents the first systematic evaluation of deep video CMC models on archival film material. We summarize representative methods and datasets, highlighting differences in model design and label definitions. Five standard video classification models are assessed on the HISTORIAN dataset, which includes expert-annotated World War II footage. The best-performing model, Video Swin Transformer, achieves 80.25% accuracy, showing strong convergence despite limited training data. Our findings highlight the challenges and potential of adapting existing models to low-quality video and motivate future work combining diverse input modalities and temporal architectures.

Problem

• Evaluating deep video models for camera movement classification in historical footage
• Assessing generalization of modern CMC methods to archival film material
• Addressing challenges of low-quality video and limited training data

Innovation

• Evaluating deep video models on historical footage
• Using Video Swin Transformer for camera movement classification
• Adapting models to low-quality archival video data