Reproduce “Emotion Transfer Using Vector-Valued Infinite Task Learning” 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
Alex Lambert, Sanjeel Parekh, Zoltán Szabó, Florence d'Alché-Buc
vITL predicts facial landmarks for requested emotions across a continuous valence–arousal space. It represents faces with 68 two-dimensional landmarks and uses kernels to encourage smooth predictions. On KDEF frontal faces with neutral input, its test MSE was 0.010 ± 0.001 versus 0.029 ± 0.003 for adapted StarGAN, supporting better landmark reconstruction in that setup, not photorealistic image generation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.