ResiComp introduces a novel framework for neural image compression that maintains quality even when data packets are lost during transmission.

• Implements dual-functional masked visual token modeling to predict missing information
• Achieves superior visual quality compared to existing methods when packet losses occur
• Enables reliable image transmission for real-time communication applications
• Balances compression efficiency with error resilience

This engineering breakthrough addresses a critical gap in neural image codecs, making them practical for deployment in unstable network environments where packet losses are common.

ResiComp: Loss-Resilient Image Compression via Dual-Functional Masked Visual Token Modeling