1 paper
Dongyoon Ryu, Sungho Jeon, Xinyue Ma +4
Edge video analytics with lightweight models is prone to accuracy degradation due to persistent distributional shifts in live video streams. While continuous learning (CL) addresse…