Reproduce “Shoggoth: Towards Efficient Edge-Cloud Collaborative Real-Time Video Inference via Adaptive Online Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Liang Wang, Kai Lu, Nan Zhang, Xiaoyang Qu, Jianzong Wang, Jiguang Wan, Guokuan Li, Jing Xiao
Shoggoth adapts edge traffic-camera object detection by running inference and training locally while a cloud teacher labels sampled frames. Replay memory and adaptive sampling help the edge model learn from new scenes while retaining earlier knowledge. On UA-DETRAC, it achieves 53.5% [email protected], versus 34.2% for Edge-Only, using 135/10 Kbps uplink/downlink; Cloud-Only scores 58.9% with 3,257/3,539 Kbps.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.