# Towards a Skeleton-Based Action Recognition For Realistic Scenarios

## TL;DR

This paper introduces a practical skeleton-based action recognition framework designed for real-world scenarios, highlighting the challenges of dataset generalization and the limitations of non-augmented data.

## Contribution

It presents a new framework for action recognition in realistic settings and discusses the impact of data normalization and augmentation on model performance.

## Key findings

- Non-augmented data performs well on test split but poorly on new datasets.
- Normalization and augmentation are crucial for real-world applicability.
- Framework demonstrates potential for service robot applications.

## Abstract

Understanding human actions is a crucial problem for service robots. However, the general trend in Action Recognition is developing and testing these systems on structured datasets. That's why this work presents a practical Skeleton-based Action Recognition framework which can be used in realistic scenarios. Our results show that although non-augmented and non-normalized data may yield comparable results on the test split of the dataset, it is far from being useful on another dataset which is a manually collected data.