VideoLT: Large-scale Long-tailed Video Recognition

TL;DR

VideoLT introduces a large-scale, long-tailed video dataset and demonstrates that existing image-based methods underperform on videos, leading to the development of FrameStack, a dynamic frame sampling technique that improves long-tailed video recognition.

Contribution

The paper presents VideoLT, a new long-tailed video dataset, and proposes FrameStack, a novel frame sampling method tailored for long-tailed video recognition.

Findings

State-of-the-art image long-tailed methods underperform on videos.

FrameStack improves classification accuracy in long-tailed video datasets.

Dynamic frame sampling balances class distribution effectively.

Abstract

Label distributions in real-world are oftentimes long-tailed and imbalanced, resulting in biased models towards dominant labels. While long-tailed recognition has been extensively studied for image classification tasks, limited effort has been made for video domain. In this paper, we introduce VideoLT, a large-scale long-tailed video recognition dataset, as a step toward real-world video recognition. Our VideoLT contains 256,218 untrimmed videos, annotated into 1,004 classes with a long-tailed distribution. Through extensive studies, we demonstrate that state-of-the-art methods used for long-tailed image recognition do not perform well in the video domain due to the additional temporal dimension in video data. This motivates us to propose FrameStack, a simple yet effective method for long-tailed video recognition task. In particular, FrameStack performs sampling at the frame-level in order to balance class distributions, and the sampling ratio is dynamically determined using knowledge derived from the network during training. Experimental results demonstrate that FrameStack can improve classification performance without sacrificing overall accuracy. Code and dataset are available at: https://github.com/17Skye17/VideoLT.