Skip to content
全プロジェクト

MLBB Video Analysis

5プラットフォーム、6バッチプロセッサー、90%以上の精度——手動審査から自動化パイプラインへ

460x faster · 1,207+ videos

5

Platforms Covered

90%+

Analysis Accuracy

1,207+

Videos Reviewed

25

Python Modules

Loading diagram...

Reviewing thousands of videos across five social platforms manually is a nightmare—inconsistent criteria, reviewer fatigue, and weeks of turnaround. We replaced that with an automated pipeline: platform-specific fetchers handle the quirks of YouTube, Instagram, TikTok, VK, and Facebook APIs, while 6 batch processors manage parallel execution with checkpointing.

The fast text analyzer achieves 90%+ accuracy at 8-thread concurrency, processing 1000+ videos per batch. For cases requiring visual verification, dedicated video analyzers extract frames and run AI inference. The system was refactored from a 25-file legacy monolith into a modular architecture with unified CLI, making it maintainable and extensible.

Python
GeminiGemini 2.5
Video AI
Analytics
Async Pipeline
PythonAI AnalysisMulti-PlatformVideo