GRACE achieves enhanced loss resilience in real-time video communication through joint training of neural encoder and decoder.
