Guardado en:
Detalles Bibliográficos
Autores principales: Ponbagavathi, Thinesh Thiyakesan, Roitberg, Alina
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2503.24298
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912916743127040
author Ponbagavathi, Thinesh Thiyakesan
Roitberg, Alina
author_facet Ponbagavathi, Thinesh Thiyakesan
Roitberg, Alina
contents Fine-grained understanding of human actions is essential for safe and intuitive human--robot interaction. We study the challenge of recognizing nearly symmetric actions, such as picking up vs. placing down a tool or opening vs. closing a drawer. These actions are common in close human-robot collaboration, yet they are rare and largely overlooked in mainstream vision frameworks. Pretrained vision foundation models (VFMs) are often adapted using probing, valued in robotics for its efficiency and low data needs, or parameter-efficient fine-tuning (PEFT), which adds temporal modeling through adapters or prompts. However, our analysis shows that probing is permutation-invariant and blind to frame order, while PEFT is prone to overfitting on smaller HRI datasets, and less practical in real-world robotics due to compute constraints. To address this, we introduce STEP (Self-attentive Temporal Embedding Probing), a lightweight extension to probing that models temporal order via frame-wise positional encodings, a global CLS token, and a simplified attention block. Compared to conventional probing, STEP improves accuracy by 4--10% on nearly symmetric actions and 6--15% overall across action recognition benchmarks in human-robot-interaction, industrial assembly, and driver assistance. Beyond probing, STEP surpasses heavier PEFT methods and even outperforms fully fine-tuned models on all three benchmarks, establishing a new state-of-the-art. Code and models will be made publicly available: https://github.com/th-nesh/STEP.
format Preprint
id arxiv_https___arxiv_org_abs_2503_24298
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Order Matters: On Parameter-Efficient Image-to-Video Probing for Recognizing Nearly Symmetric Actions
Ponbagavathi, Thinesh Thiyakesan
Roitberg, Alina
Computer Vision and Pattern Recognition
Fine-grained understanding of human actions is essential for safe and intuitive human--robot interaction. We study the challenge of recognizing nearly symmetric actions, such as picking up vs. placing down a tool or opening vs. closing a drawer. These actions are common in close human-robot collaboration, yet they are rare and largely overlooked in mainstream vision frameworks. Pretrained vision foundation models (VFMs) are often adapted using probing, valued in robotics for its efficiency and low data needs, or parameter-efficient fine-tuning (PEFT), which adds temporal modeling through adapters or prompts. However, our analysis shows that probing is permutation-invariant and blind to frame order, while PEFT is prone to overfitting on smaller HRI datasets, and less practical in real-world robotics due to compute constraints. To address this, we introduce STEP (Self-attentive Temporal Embedding Probing), a lightweight extension to probing that models temporal order via frame-wise positional encodings, a global CLS token, and a simplified attention block. Compared to conventional probing, STEP improves accuracy by 4--10% on nearly symmetric actions and 6--15% overall across action recognition benchmarks in human-robot-interaction, industrial assembly, and driver assistance. Beyond probing, STEP surpasses heavier PEFT methods and even outperforms fully fine-tuned models on all three benchmarks, establishing a new state-of-the-art. Code and models will be made publicly available: https://github.com/th-nesh/STEP.
title Order Matters: On Parameter-Efficient Image-to-Video Probing for Recognizing Nearly Symmetric Actions
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2503.24298