Reproduce “Multi-face emotion detection for effective Human-Robot Interaction” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Mohamed Ala Yahyaoui, Mouaad Oujabour, Leila Ben Letaifa, Amine Bohi

A system for real-time multi-face emotion detection was developed and deployed on a mobile humanoid robot, addressing the challenge of integrating computationally intensive models onto resource-constrained platforms. The research balanced high accuracy (70.00% with EfficientNetV2-B0) with a compact memory footprint (139 MB) to enable dynamic emotion display and adaptive robotic behavior for multiple human interactants.

Key insights

  • Achieving effective HRI with emotion detection on mobile robots necessitates a careful trade-off between a deep learning model's accuracy and its memory footprint and computational demands.
  • Confidence interval analysis is a valuable tool for statistically comparing models and selecting the most robust option when multiple models exhibit similar nominal performance but varying resource requirements.
  • Even with advanced deep learning for emotion recognition, simpler, real-time face detection methods like Haarcascade can effectively anchor multi-face systems on resource-constrained platforms.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.