Reproduce “MEGC2025: Micro-Expression Grand Challenge on Spot Then Recognize and Visual Question Answering” 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
Xinqi Fan, Jingting Li, John See, Moi Hoon Yap, Wen-Huang Cheng, Xiaobai Li, Xiaopeng Hong, Su-Jing Wang and 1 more
The MEGC 2025 Grand Challenge advanced micro-expression (ME) analysis by introducing an integrated Spot-Then-Recognize (ME-STR) pipeline and a novel ME Visual Question Answering (ME-VQA) task utilizing large vision-language models. Top teams achieved an ME-STR score of 0.09 (vs. baseline 0.0062-0.0086) and an ME-VQA average score of 0.575 (vs. Qwen2.5VL-3B baseline), demonstrating progress in both practical integration and interpretable ME understanding.
Key insights
- ME spotting remains a significant bottleneck in the integrated ME-STR pipeline, often limiting the overall performance despite improvements in recognition.
- The ME-VQA task effectively leverages MLLMs/LVLMs to enable more interpretable and nuanced understanding of micro-expressions through natural language interactions.
- Explicit temporal and motion information, such as onset, apex, offset (OAO) frames and optical flow, improves emotion classification in ME-VQA tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.