Search papers, labs, and topics across Lattice.
Mary University of London
2
0
2
Task-Adaptive Pruning (TAP) achieves a remarkable 47.0 mIoU on ADE20K and 53.7 box AP on COCO, all while enhancing encoder throughput by over 30%.
Achieving high-fidelity avatar animation with just a single frame, SAGE revolutionizes the data requirements for dynamic facial expression modeling.
