1 views collected around this technical thread.
This article presents Meitu Community's AI‑driven multimedia content analysis pipeline, covering short‑video classification, video fingerprinting, and OCR, detailing model choices, experimental results, and future directions for improving content audit, quality, tagging, and feature engineering.
Meitu replaced on-device beautification and video processing with a cloud-native architecture that routes requests by region, uses a dedicated upload SDK for detailed monitoring, employs edge-computing, a configuration-driven plug-in framework and Kubernetes-based elastic scaling, enabling fast, reliable, globally-distributed image and video services.
In the 2018 AI Challenger, Meitu co‑organized a new “Real‑Time Short‑Video Classification” track and released the industry’s first multi‑label short‑video dataset of 200,000 mobile‑captured, vertically oriented videos spanning 63 categories and detailed tags for subjects, scenes, actions, and other dimensions, advancing video semantic understanding and AI research.
Meitu Live’s hybrid‑cloud architecture combines a proprietary push‑stream SDK, custom source servers, and an in‑house player with big‑data‑driven quality monitoring, multi‑CDN integration, and public‑cloud nodes, enabling over 100,000 daily sessions while sharing practical optimization lessons and system‑building insights.
At Meitu’s third Internet Technology Salon in Xiamen on November 26 2016, over 150 senior engineers heard Meitu’s Lu Rongbin detail the company’s progression from simple rsync scripts to a scalable mobile data and open statistical platform, while Tencent’s Zhao Shiyuan showcased the Glacier multi‑dimensional analysis system for fast, tag‑driven queries, underscoring collaborative technical exchange in South China.
Since mid‑2015 Meitu has been integrating Golang to build high‑performance, highly‑available online services, creating supporting components and infrastructure, and after a year of production use the team identified numerous pitfalls, gathered practical insights, and now shares deployment challenges, architectural decisions, and best‑practice recommendations.
The Meitu Technology Salon’s second offline session, held on July 24 2016, gathered senior R&D engineers to explore practical Golang applications through three case studies—Meitu’s high‑performance services, Panda TV’s live‑streaming architecture overhaul, and NetEase’s P2P cache system redesign—offering insights and networking for 100 participants.