Label Studio

Adoption assessment: Label Studio is frequently recommended in developer communities like Zhihu, with numerous tutorials ranging from basic setup to advanced ML-backend integration. This indicates strong awareness among beginners and experimenters, but deep feedback from large-scale production use remains sparse. What it does: Label Studio handles labeling for images, text, audio, video, and time-series data, offering interfaces for bounding boxes, polygons, keypoints, NER, text classification, speech transcription, etc. It can integrate with ML backends (e.g., Qwen2-VL, YOLO) for automatic pre-annotation, drastically improving efficiency, and supports exporting to various model formats. Barriers and cost: The community edition is free, easily installed via pip or Docker, and is great for learning and solo projects. However, it lacks built-in multi-user collaboration. Team-based annotation requires the Enterprise edition, which according to community-collected data (not official) starts at $99/month + $49/user/month for up to 12 users.