Wireless LANs have become an essential part of modern communication infrastructure. However, their performance often suffers from physical obstructions and traffic congestion, leading to unstable channel conditions. This instability can severely impact latency-sensitive applications such as video streaming. Conventional systems typically transmit data without considering network conditions or the semantic importance of content, risking the loss of critical information. To address this issue, we propose an adaptive frame transmission method that selectively sends video data based on wireless conditions and content-level importance. An importance value, calculated from factors such as confidence scores and motion, is assigned to each video frame or individual detected object based on object detection analysis. The system estimates available throughput and transmits data units only if their predicted delay is within the predetermined acceptable threshold. We implement this method over the QUIC protocol, utilizing its application-layer control and support for unreliable datagrams. Preliminary evaluations, conducted using MJPEG, show that our approach reduces transmitted data by 71% and average delay by 47%, while preserving important content.