Label Studio

Label Studio is an open source platform for data labeling, AI evaluation and human in the loop workflows. It supports many data types in one tool, including images, video, audio, text, documents, time series and multi modal data, and is configured through customizable labeling interfaces. Typical jobs include computer vision work such as image classification, object detection, object tracking and semantic segmentation, with ML models available to pre label data. For language and documents it covers named entity recognition, question answering, sentiment analysis and PDF and image OCR annotation. For audio it supports transcription, speaker diarization and emotion recognition. It also handles AI evaluation tasks: collecting human preferences and rankings for RLHF and fine tuning, building custom benchmarks and rubrics for side by side LLM comparison, grading RAG answers against source material, and connecting observability tools so people can review agent traces. A recent addition is video object segmentation and tracking using SAM 2. It suits machine learning engineers, data scientists and annotation teams that want to run labeling themselves. The software can be installed with pip, Homebrew or Docker, or run from the Git repository, and then launched locally. The open source edition is free to download. The site also offers a comparison of versions for those who want more than the open source edition.