Reproduce “Describe Your Facial Expressions by Linking Image Encoders and Large Language Models” 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
Yujian Yuan, Jiabei Zeng, Shiguang Shan
Exp-BLIP links visible facial action cues to inferred emotion by training on text built from existing labels. Separate action and emotion captioners provide descriptions that GPT-3.5 fuses. In a human evaluation of ViT-B/OPT-2.7B models, Exp-BLIP was chosen most complete 67.72% of the time, versus 28.85% for direct concatenation and 3.43% for mixed captions; this reflects relative preference, not proof of a person’s true inner emotion.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.