In the Internet of Vehicles (IoV), data-intensive sensing and perception tasks require the transmission of large

amounts of information, while available bandwidth of the whole system is often limited. To address this issue, this paper proposes a semantic-driven Communication framework which prioritizes task-relevant semantic information and transmitting semantic labels for contextual regions, significantly reducing bandwidth consumption while preserving semantic integrity. At the receiver, a GAN-based reconstruction module is further designed to recover contextual information from the received semantic cues, ensuring that task-related semantics are preserved. Experimental results show that the proposed framework can effectively reduce

bandwidth consumption while maintaining high semantic reconstruction quality and strong robustness, particularly under low signal to interference plus noise ratio (SINR) conditions.