Search papers, labs, and topics across Lattice.
Intelligent Interconnected Systems Laboratory of Anhui Province
2
0
3
A learnable missing token representation allows MRAF to achieve 100% accuracy in polyglot speaker identification, even when face information is missing.
Standard multimodal fusion can hurt performance in emotion recognition, but this new approach knows when to drop modalities, leading to state-of-the-art results.
