Frame2Freq: Spectral Adapters for Fine-Grained Video Understanding

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ponbagavathi, Thinesh Thiyakesan, Seibold, Constantin, Roitberg, Alina
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917286592380928
author Ponbagavathi, Thinesh Thiyakesan
Seibold, Constantin
Roitberg, Alina
author_facet Ponbagavathi, Thinesh Thiyakesan
Seibold, Constantin
Roitberg, Alina
contents Adapting image-pretrained backbones to video typically relies on time-domain adapters tuned to a single temporal scale. Our experiments show that these modules pick up static image cues and very fast flicker changes, while overlooking medium-speed motion. Capturing dynamics across multiple time-scales is, however, crucial for fine-grained temporal analysis (i.e., opening vs. closing bottle). To address this, we introduce Frame2Freq -- a family of frequency-aware adapters that perform spectral encoding during image-to-video adaptation of pretrained Vision Foundation Models (VFMs), improving fine-grained action recognition. Frame2Freq uses Fast Fourier Transform (FFT) along time and learns frequency-band specific embeddings that adaptively highlight the most discriminative frequency ranges. Across five fine-grained activity recognition datasets, Frame2Freq outperforms prior PEFT methods and even surpasses fully fine-tuned models on four of them. These results provide encouraging evidence that frequency analysis methods are a powerful tool for modeling temporal dynamics in image-to-video transfer. Code is available at https://github.com/th-nesh/Frame2Freq.
format Preprint
id arxiv_https___arxiv_org_abs_2602_18977
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Frame2Freq: Spectral Adapters for Fine-Grained Video Understanding
Ponbagavathi, Thinesh Thiyakesan
Seibold, Constantin
Roitberg, Alina
Computer Vision and Pattern Recognition
Adapting image-pretrained backbones to video typically relies on time-domain adapters tuned to a single temporal scale. Our experiments show that these modules pick up static image cues and very fast flicker changes, while overlooking medium-speed motion. Capturing dynamics across multiple time-scales is, however, crucial for fine-grained temporal analysis (i.e., opening vs. closing bottle). To address this, we introduce Frame2Freq -- a family of frequency-aware adapters that perform spectral encoding during image-to-video adaptation of pretrained Vision Foundation Models (VFMs), improving fine-grained action recognition. Frame2Freq uses Fast Fourier Transform (FFT) along time and learns frequency-band specific embeddings that adaptively highlight the most discriminative frequency ranges. Across five fine-grained activity recognition datasets, Frame2Freq outperforms prior PEFT methods and even surpasses fully fine-tuned models on four of them. These results provide encouraging evidence that frequency analysis methods are a powerful tool for modeling temporal dynamics in image-to-video transfer. Code is available at https://github.com/th-nesh/Frame2Freq.
title Frame2Freq: Spectral Adapters for Fine-Grained Video Understanding
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2602.18977