NoisyActions2M: A Multimedia Dataset for Video Understanding from Noisy Labels

TL;DR

This paper introduces NoisyActions2M, a large-scale multimedia dataset with noisy user-generated labels for advancing video understanding, and demonstrates its utility in training robust models and improving performance on downstream tasks.

Contribution

The paper presents a new large-scale noisy video dataset, along with analysis of training strategies and its effectiveness in enhancing video classification models.

Findings

Pretraining on NoisyActions2M improves performance on UCF101 and HMDB51.

The dataset helps models become more robust to video corruption and label noise.

Different loss functions and pretraining strategies are evaluated for noisy video learning.

Abstract

Deep learning has shown remarkable progress in a wide range of problems. However, efficient training of such models requires large-scale datasets, and getting annotations for such datasets can be challenging and costly. In this work, we explore the use of user-generated freely available labels from web videos for video understanding. We create a benchmark dataset consisting of around 2 million videos with associated user-generated annotations and other meta information. We utilize the collected dataset for action classification and demonstrate its usefulness with existing small-scale annotated datasets, UCF101 and HMDB51. We study different loss functions and two pretraining strategies, simple and self-supervised learning. We also show how a network pretrained on the proposed dataset can help against video corruption and label noise in downstream datasets. We present this as a benchmark dataset in noisy learning for video understanding. The dataset, code, and trained models will be publicly available for future research.