Label Studio
summary
Label Studio is a web-based annotation platform that claims to accelerate dataset creation by providing a web-based annotator with configurable tasks: image bounding boxes and segmentation, text classification and sequence labeling, audio transcription and time spans. It installs via Docker or pip, exposes a REST API and Python SDK, exports to COCO/YOLO/VOC/CSV, supports multi-user projects and versioned label histories, and enables model-in-the-loop workflows so trained models prelabel data for humans to correct. Practical demos show interactive UI and export pipelines that plug into typical ML training loops.
AI-written summary · about
related talks
40mData labelling to scale your ML and AI pipelines
59mWorkshop: Choosing the Best Open Source LLM for Your Application
31mCreate text classifiers in a few hours using the open-source Label Sleuth system
39mActive learning in the interactive python environment
36mProgrammatically creating and managing training data
21mAWS On Air ft. Amazon SageMaker Ground Truth Plus
16mRight data at the right time: Tools demo
30mEfficient data labelling with weak supervision