Cross-Model Pseudo-Labeling for Semi-Supervised Action Recognition

TL;DR

This paper introduces Cross-Model Pseudo-Labeling (CMPL), a novel semi-supervised action recognition method that leverages two models predicting for each other to improve learning from limited labeled data.

Contribution

The work proposes a new pseudo-labeling scheme with two models that learn from each other's predictions, enhancing semi-supervised action recognition performance.

Findings

CMPL achieves 17.6% and 25.1% Top-1 accuracy on Kinetics-400 and UCF-101 with only 1% labeled data.

CMPL outperforms baseline FixMatch by 9.0% and 10.3% on the respective datasets.

Experiments show significant improvements over existing semi-supervised methods.

Abstract

Semi-supervised action recognition is a challenging but important task due to the high cost of data annotation. A common approach to this problem is to assign unlabeled data with pseudo-labels, which are then used as additional supervision in training. Typically in recent work, the pseudo-labels are obtained by training a model on the labeled data, and then using confident predictions from the model to teach itself. In this work, we propose a more effective pseudo-labeling scheme, called Cross-Model Pseudo-Labeling (CMPL). Concretely, we introduce a lightweight auxiliary network in addition to the primary backbone, and ask them to predict pseudo-labels for each other. We observe that, due to their different structural biases, these two models tend to learn complementary representations from the same video clips. Each model can thus benefit from its counterpart by utilizing cross-model predictions as supervision. Experiments on different data partition protocols demonstrate the significant improvement of our framework over existing alternatives. For example, CMPL achieves $17.6\%$ and $25.1\%$ Top-1 accuracy on Kinetics-400 and UCF-101 using only the RGB modality and $1\%$ labeled data, outperforming our baseline model, FixMatch, by $9.0\%$ and $10.3\%$, respectively.