MPL: Mine the pseudo labels for semi-supervised temporal action localization

Key Points

• Improved performance in temporal action localization was observed, demonstrating the effectiveness of pseudo labels.
• Models utilizing pseudo labels achieved a 15% increase in accuracy compared to those without them.
• Analysis of video data with semi-supervised learning techniques highlights the role of pseudo labels.
• These findings suggest that integrating pseudo labels may enable more robust action localization in machine learning.